TDOS6.WS4
---------
- "TurboDOS spans the Horizon"
  Karl Sterne, North Star Computers, Inc.
  "Microsystems", August 1984, Vol.5, No.8, p.114
(Retyped by Emmanuel ROCHE.)

The  North  Star  Horizon,  now  6 years old,  is  one  of  the  few  pre-1980
microcomputers still in demand. However, the Horizon, as it is built in 1984,
is a very different system from its 1978 ancestor. Today's Horizon is a multi-
user  multi-processor  that uses North Star's version  of  TurboDOS  operating
system  from  Software  2000, Inc., for both 8-bit  CP/M  and  16-bit  CP/M-86
applications.
By  the end of 1982, it was apparent that the Horizon's lifespan would be  far 
longer  than  had  been forecast. Its chief limitation  was  its  time-sharing
operating  system. To replace it, North Star evaluated a number  of  operating
systems before choosing TurboDOS. Among the reasons for the decision were:
     1. TurboDOS  has  a multi-processor networking  architecture.  (In  fact, 
        North   Star's  latest  computer,  the  Dimension,  uses   a   similar
        architecture. The Dimension, however, incorporates a multiple-user IBM
        XT-compatible  operating system rather than TurboDOS, as well  as  the
        Intel  80186  processor rather than the Zilog Z-80, and  the  IBM  bus
        rather than the S-100 Bus.)
     2. TurboDOS  provides true multi-user operation, including  file  sharing 
        and record lockout.
     3. TurboDOS   permits   simultaneous  operation  of  8-bit   and   16-bit 
        applications.
     4. TurboDOS outperforms many other operating systems in timed benchmarks.
     5. TurboDOS  has  powerful networking facilities,  including  support  of 
        multiple circuits, node-to-node communications, and the ability of any
        node to be a server. This allows, for example, direct access to the S-
        100 Bus for control of special boards.
     6. TurboDOS comes as close to being a minicomputer operating system as is 
        practical on a microcomputer.
Given  all its advantages, TurboDOS can still be improved. This  article  will 
discuss some of the improvements made by North Star.

16-bit operation
----------------
North Star's principal problem with TurboDOS was that Release 1.22, the latest 
available  in  1983,  emulated  only  8-bit CP/M,  and  did  not  have  16-bit
capabilities.  Software  2000's assurance, that a  new  release  incorporating
CP/M-86  emulation  was  imminent,  allowed North Star  to  proceed  with  the
development  of  new hardware: an 8-bit Zilog Z-80 satellite board,  a  16-bit
Intel  8088-2 satellite board, and a memory expansion board to give the  Intel
8088-2 as much as 512K of RAM. The existing Horizon hardware was used  without
modification  for the server (master). Although both the 16-bit  hardware  and
the  new  TurboDOS release took longer than expected, they have  been  in  use
since April 1984, and have had excellent acceptance.

Installation and configuration
------------------------------
TurboDOS  runs  on  a  large  variety of  hardware,  and  one  result  of  its 
versatility is that it tends to be difficult to install. Having once succeeded
in  installing the "vanilla" system, the user is confronted with the  task  of
tailoring  it  to  that particular hardware. This  usually  requires  a  word-
processor to create lists of operating system functions and other  parameters.
This  process is a radical departure from the straightforward installation  of
other North Star operating systems.
The  North  Star  approach  was  to  split  the  process  into  2   functions: 
configuration  and  installation.  The configuration  program  asks  the  user
questions  in English, and does not require a word-processor; installation  is
automated, using the operator's responses.
The  North Star TurboDOS operating system is distributed as a 4-diskette  set. 
Each  diskette  is  named  for its particular  function  in  the  installation
process:  SYSTEM  DISK, CONFIG DISK, HELP DISK, and SYSCON  DISK.  The  SYSTEM
DISKette  is  a  bootable TurboDOS operating system that  contains  a  maximum
hardware  configuration.  Some  users  will never  need  to  generate  another
operating system.
Two  manuals come with North Star TurboDOS: the TurboDOS User's Guide and  the 
TurboDOS  Reference Manual. The preface to the TurboDOS User's Guide  contains
step-by-step instructions for helping the system install itself. Three  simple
commands are entered from any user's console. The rest of the process requires
only a few Carriage Returns and diskette changes.
The  installation process uses the TurboDOS command file batch utility  --  an 
enhancement   of  the  CP/M  SUBMIT  facility  --  and  performs  the   system
initialization  tasks  such  as  verifying  the  hard  disk  data  tracks  and
formatting  the TurboDOS directory area. The user is then asked to insert  one
of  the 4 distribution diskettes. The proper user areas of the hard  disk  are
loaded with the appropriate files. This process is repeated for each of the  4
diskettes.
At this point, the installation creates a bootable diskette for daily use. The 
distribution  diskette  set  can  be set aside,  and  the  computer  is  fully
operational.  If  customizing  of the operating system is  not  desired,  this
completes the process.
The  operating  system  on the SYSTEM DISK contains  software  drivers  for  a 
maximum  system.  It includes hard disk drivers for both types of  North  Star
hard  disks, and 2 different kinds of printer drivers. Should the user wish  a
different configuration, the CONFIG program is run.
CONFIG  asks the user questions in simple English about the  desired  hardware 
configuration. It builds the TurboDOS GENeration and PARameter files  required
by  the TurboDOS GEN command. No other programs are required. After  the  user
finishes  answering  all the questions, a system summary is displayed  on  the
screen.  This can be accepted or aborted, and the user can change any  desired
parameters.
At the end of the session, the user can opt not to have the TurboDOS operating 
system  actually  generated.  In this mode, CONFIG acts as  a  teaching  tool,
allowing  the user to see how different configurations change the form of  the
GEN and PAR files.
If  additional  TurboDOS operating systems are desired, another  command  file 
batch  is  executed.  This  file, created by  CONFIG,  performs  the  TurboDOS
operating  system generation and copies the new TurboDOS operating systems  to
the proper area of the hard disk. The old TurboDOS operating system files  are
saved with an ORG filetype. Should any problems occur with the newly-generated
TurboDOS operating systems, the old ones can be recovered.

Bad spot de-allocation
----------------------
All  hard disk systems must deal with the question of how to detect and  avoid 
defects (known as bad spots) on the hard disk medium. Typically, a disk  drive
will  be shipped by the manufacturer with a few bad spots already on  it,  and
additional bad spots will "grow" as the result of vibration (especially during
shipping), power failures, or aging of the machine.
A  good  hard  disk system must, therefore, deal with  2  different  bad  spot 
scenarios:
        1. A hard disk arrives with bad spots already on it.
        2. A hard disk grows bad spots while it is in use.
Hard disk systems must also deal with 2 kinds of bad spots: "hard" (permanent) 
bad spots and "soft" (intermittent) ones. It is good practice to recognize and
avoid the soft ones as well as the hard ones (even though you can often  retry
enough  times to get past the soft ones) because they tend to get  worse  with
age.
The  issue  is complicated by the fact that the location of the bad  spot  can 
make a big difference in how bad it really is. A bad spot in an unused part of
the disk is not a problem, as long as one can tell the operating system how to
avoid it. A bad spot in a data area is a problem, because data has very likely
been lost. A bad spot in the directory can be fatal.
Generic  TurboDOS comes with a program, VERIFY, which de-allocates bad  spots; 
i.e.,  it  removes them from the pool of available disk space.  VERIFY  has  2
deficiencies, however. First, it finds only hard errors, because it is  forced
to  do  read-only  tests via the normal hard disk drivers,  which  are  fault-
tolerant by design. Soft errors will trigger retries at the driver level,  but
these  retries  are  not reported to VERIFY unless  many  successive  failures
occur.  Second, it can be run only at startup -- the directory must  be  empty
for  it  to work properly. Therefore, it does not help at all with  bad  spots
that grow during use.
North Star's answer was to create a program named MARKBAD. MARKBAD is  similar 
to  VERIFY  in  that it de-allocates disk blocks that contain  bad  spots.  It
differs from VERIFY in 2 important ways. First, it accepts manual input of bad
spots. This allows identification of both soft and hard spots, which are taken
from  the  manufacturer's disk label, from a hard disk test program,  or  from
disk  error messages put out by TurboDOS itself. Second, it can be run at  any
time,  so  that  a  bad spot that grows during use can  be  removed  from  the
available pool.
VERIFY  and MARKBAD both deal with bad spots in the disk's data area.  Neither 
can help if the bad spot is in the directory, because directory blocks  cannot
be  de-allocated.  TurboDOS  requires the  entire  directory  area  (including
allocation  table) to be free of defects. On a 30MB hard disk with a 2K  block
size,  for example, this area occupies 30 tracks, or about 240K.  (ROCHE>  The
size of a standard IBM 3740 8" floppy disk...)
To  alleviate  this situation, North Star developed a means  of  swapping  bad 
directory  tracks with good data tracks in a manner invisible to TurboDOS,  so
that the bad blocks end up in the data area (where they can be de-allocated by
MARKBAD)  and  the  good blocks end up in the directory.  This  preserves  the
maximum  amount of good disk space possible. Another approach would have  been
to  slide  the beginning of the directory out into the first clear  space  big
enough  to  hold it, but a potentially large amount of good disk  space  might
have to be skipped, and that space would be lost.
The swapping of tracks takes place on power up, when a special section of  the 
hard  disk driver reads the North Star bad-spot table from a reserved  portion
of the disk. This bad-spot table is initially written in the factory, and  can
be updated in the field by running the hard disk test-and-format program.
When a disk is shipped with a bad spot in the directory, the first system boot 
will  swap the bad track out into the data area, and MARKBAD will be  told  to
de-allocate  the  affected data blocks. The system will then appear  like  any
other North Star TurboDOS system.
If  a bad spot grows in the directory later, some or all of the disk  will  be 
unreadable.  The procedure is to recover what can be read, then run  the  hard
disk  test-and-format program, and tell it where the new bad spot is.  On  the
next  boot,  the new bad track will be swapped out of the directory,  and  the
system will again be usable. Any lost data has to be recovered from the backup
media.

User interface
--------------
Even though TurboDOS provides a considerably more pleasant user interface than 
CP/M,  it  is designed for computer professionals rather than  for  the  small
businesses that are North Star's principal customers. To present a more easily
understood  set of screens, North Star has bundled Turbo-Plus into North  Star
TurboDOS.  This  is an enhancement package that provides  powerful  additional
facilities  for  TurboDOS. The utilities included are:  DIRDUMP  displays  the
master  directory of any disk; WHO displays a list of all the  current  users;
LOCATE searches any or all drives for a file; BB (Background Batch)  schedules
jobs  to  the  background queue; STATUS monitors the  activity  of  users  and
peripherals;  HELP provides on-line help menus that users may  customize;  TWX
(ROCHE> TeletypeWriter eXchange. An old US and Canadian dial-up communications
service that became part of Telex.) sends messages to other users immediately;
MAIL is an electronic mail facility.
Turbo-Plus is a set of utilities developed by Microserve Inc. for TurboDOS. It 
was  chosen primarily for its extensive on-line HELP messages, as well as  for
its  versatile  electronic mail facility. In addition, Turbo-Plus  contains  a
group of commands that allows the network manager to track utilization, keep a
log of user time, and control other users.
Besides  these  aids  for  less sophisticated users,  Turbo-Plus  also  has  a 
powerful  Background  Batch utility, BB, that allows users  to  schedule  low-
priority  non-interactive  jobs for execution in background mode or  at  times
when the system is lightly used.

Conclusion
----------
By  adding TurboDOS and the new multi-processor hardware developed for  it  to 
the  Horizon, North Star has extended the usage of this popular  computer  for
years to come. And North Star's implementation of TurboDOS brings the power of
a  sophisticated operating system to non-professional users who need  only  to
follow a step-by-step procedure for successful installation and operation.

EOF
TDOS8.WS4
---------
- "For Networks and Multi-user Systems -- TurboDOS"
  Ron Fowler
  "Microsystems", Vol.5, No.8, August 1984, p.58
(Retyped by Emmanuel ROCHE.)

Gone  are the days when simple operating systems satisfied the needs  of  most
small-system  users. Once written primarily to satisfy the  relatively  modest
requirements  of hobbyists and experimenters, microcomputer operating  systems
are  now designed for performance, with special concern for providing  maximum
versatility to the user.
One  such  operating  system is TurboDOS, a product  of  Software  2000,  Inc. 
Written  originally for the Zilog Z-80 (and recently converted for  the  Intel
8088/8086  family), this package offers a high degree of  CP/M  compatibility,
multi-tasking background processing, powerful networking capability,  built-in
communications facilities, print spooling, and much more.
This article will take an in-depth look at this powerful system: what you  can 
expect  to  gain by using TurboDOS, and what (if anything) you might  lose  by
giving  up CP/M in favor of TurboDOS. We will examine the  TurboDOS  operating
system  and its utilities in detail, and conclude with a section addressed  to
the   programmer   or  system  developer  interested   in   writing   TurboDOS
applications.
(ROCHE> Unfortunately, when I read this article, 30 years later, it is obvious 
that  Ron  Fowler  had no experience of MP/M,  the  multi-tasking,  multi-user
version  of  CP/M...  By  the way, each time that he  says  "CP/M",  you  must
understand:  "CP/M  2.2". TurboDOS is, in fact, a multi-processor  version  of
MP/M.  That's why it is more compatible with MP/M than with CP/M  (2.2)!  Poor
Ron  Fowler,  not  knowing  MP/M, is often ecstatic  by  the  capabilities  of
TurboDOS, which are exactly (except multiple processors) those of MP/M...)

Background
----------
Much   of  the  information  presented  here  is  based  on  my   own   custom 
implementation  of TurboDOS Release 1.3. The evaluation copy of  the  software
was provided courtesy of North Star Computers, Inc. North Star provides a full
implementation for their own hardware, and offers it bundled with  Turbo-Plus,
a TurboDOS enhancement package (which I hope to review in a future article).
While  the  1.3 Release of TurboDOS provides Intel  8088  compatibility,  this 
discussion  is limited to the Zilog Z-80 version. The evaluation system  is  a
networked  version,  interconnecting  a Z-80-based S-100  Bus  system  with  a
disk-less  single-board  Z-80 computer, through Intel 8255 PIO chips  in  each
computer.  Figure 4 is a schematic of the circuit used to interconnect  the  2
systems.
The  S-100  Bus  system consists of an Integrand  mainframe  housing  a  Delta 
Products 4-MHz Z-80 CPU, 128K of CompuPro memory, and various I/O boards.

Overview
--------
TurboDOS  is  structured in a 2/3 arrangement: the main body of  the  TurboDOS 
operating system is analogous to CP/M's CCP and BDOS. This portion is machine-
independent,  and is supplied as a group of Microsoft-format REL  (relocatable
machine  code) files. The remainder of the system is similar to  CP/M's  BIOS,
and  consists  of a collection of *DEVICE DRIVERS* (for such  things  as  disk
drives,  printers,  and  network communications).  These  driver  modules  are
written  by  the system integrator; if you purchase a packaged  system,  these
drivers are already written for you.
The modular design of the TurboDOS operating system allows it to be configured 
in  an  almost infinite variety of forms. Table 1 lists a number of  the  more
commonly-used modules.
Table 1. Common TurboDOS modules
(NOTE: Modules denoted by an asterisk (*) after the module names are  resident 
processes.)
Name    Function
----    --------
LCLUSR* Supports the Transient Program Area
LCLMSG  Contains the system error messages
CMDINT  Command line interpreter, similar to CP/M's CCP
AUTLOG  Supports COLDSTRT.AUT and WARMSTRT.AUT files
NETSVC* Services  network requests. Necessary only in processors that  provide
        facilities to the network (servers)
NETREQ  Makes requests of the network
NETTBL  Defines network topology from viewpoint of the local node
DSPOOL* Printer de-spooler (background print)
FLUSHR* Periodically flushes disk buffers
OSNTRY  Decodes system calls (C-functions and T-functions)
FILMGR  Handles the local file system.
        Nodes without disk storage do not require this module.
FILLOK  Handles file interlocks
FFOMGR  Handles the special FIFO files
FASLOD  Fast program-load optimizer
BUFMGR  Manages the disk buffers
DSKTBL  Defines the available disk drives (local and remote)
CPMSUP  Special support for CP/M.
        Makes TurboDOS operating system more CP/M-compatible.
CONMGR  Handles all console I/O
CONTBL  Connects CONMGR to hardware console driver module(s)
DOMGR   Handles batch commands
LSTMGR  Handles printer output
SPOOLR  Handles background spooled-print functions, as well as routine printer
        output over the network
RTCMGR  Maintains system date and time (real-time clock)
BNKMGR  Banked memory support
DSPCHR  Multi-tasking  kernel;  controls  CPU  dispatching  between  competing
        processes, as well as synchronization and inter-process communications
DSPSGL  Special "null" dispatcher; allows a non-multi-task version of TurboDOS
        to be created
MEMMGR  Controls dynamic allocation of memory
OSLOAD  Special module used to create a loader program
COMSUB  Common subroutines needed in all versions of TurboDOS
Each major function is isolated in its own module, many of which can be  added 
to  and  deleted from the TurboDOS operating system by the user  (ROCHE>  ?  A
*USER*  fine tuning a Disk Operating System?), thus providing the  ability  to
add  or  subtract functionality. For example, the SPOOLR module  contains  the
necessary  code to implement the TurboDOS print spooling  function  (described
later). By deleting this module in systems that do not require print  spooling
(e.g.,  single-user  systems),  the  TurboDOS operating  system  can  be  made
smaller, allowing more room for disk buffers and transient programs.
Flexible as this assortment of modules may be, it would be demanding a lot  to 
ask  all  users to build their own TurboDOS operating system. To  avoid  this,
TurboDOS  is  packaged with a number of "standard" versions  (Table  2),  each
providing a commonly-required configuration.
Table 2. Standard TurboDOS configuration packages
Name      Function
----      --------
STDLOADR  Cold-start system loader
STDSINGL  Single-user version, no spooling
STDSPOOL  Single-user version with spooling
STDMASTR  Network master
STDSLAVE  Network slave without local disks
STDSLAVX  Network slave with local disks
(ROCHE>  So, The first 3 versions correspond to CP/M, and the last 3  versions 
to MP/M. Note the emphasis on spooling. In fact, Digital Research was  selling
one  DESPOOL  program for CP/M Versions 1 and 2, but it only ran  reliably  on
DRI's Intel MDS-800 microcomputers...)
The system may be configured (and reconfigured) conveniently by using GEN.COM, 
the  supplied  system  generation  utility.  This  program  is,  in  fact,   a
sophisticated  linkage  editor  that  takes  its  commands  from  disk  files,
providing  a  high degree of automation in the system generation  process.  An
additional file (called a PAR file) can be specified to GEN.COM. The PAR  file
references, *BY NAME*, parameters within the system. It is a powerful concept:
you  can  restructure  major  portions of the  system  (such  as  the  network
definition  tables,  disk assignment tables, command-search  paths,  and  many
others) merely by editing the PAR file and regenerating the system.
Most implementations of TurboDOS provide for a boot PROM, allowing the  system 
to start up automatically at power-on. The boot PROM contains just enough code
to  load a file named OSLOAD.COM into the program area. OSLOAD then brings  in
the  rest of the TurboDOS operating system. My own implementation  boots  from
CP/M,  running  OSLOAD as a CP/M transient program. This suits  my  particular
needs better than setting up a boot PROM -- and, incidentally, illustrates the
versatility of TurboDOS.
When  the  TurboDOS operating system start up, it prints  a  sign-on  message, 
followed  by the system prompt. This is where TurboDOS begins to  differ  from
standard CP/M: the prompt contains the user number, as well as the drive name.
(TurboDOS supports 32 user numbers -- twice as many as CP/M; and the  familiar
angle  bracket  prompt  character of CP/M is replaced with  a  curly  bracket:
"A}".)
Depending on how the system is set up, it may be necessary to *LOG IN*  before 
any useful work can begin. A utility, LOGON.COM, is provided for this purpose,
and  another,  LOGOFF.COM,  is used to terminate the  session.  TurboDOS  may,
optionally,  be  instructed  to maintain a file of  logons  and  logoffs,  and
require  that a password be specified at login. Passwords are maintained in  a
special user accounts file maintained with a text editor. Additional  security
allows  logon  to  be retricted to a single user  area  (this  restriction  is
specified  in  the  user accounts file). Another  login  option  provides  for
privileged  logins  (non-privileged  logins may not  change  user  areas,  and
several of the supplied utilities will function only for privileged logins).
The  rhythm of this system is a bit different from that of CP/M. For  example, 
there  is no warm boot, since the entire TurboDOS operating system remains  in
memory  at  all  times.  (CP/M 2.2, of  course,  allows  the  Console  Command
Processor  to be overwritten by a user program, thus requiring that a  portion
of  the  CP/M 2.2 operating system be read back in from disk  when  a  program
terminates) (ROCHE> CP/M Plus ("CP/M Version 3" in the USA), launched in  1983
(so,  one year before this article), is a single-user version of  MP/M...  So,
since it supports "banked memory", the CCP is reloaded instantly from  memory.
Each  time CP/M Version 2 was spending too much time reading the floppy  disk,
the  data was stored in memory. 2 examples: the CCP and the directory.  That's
why CP/M Plus, in a good implementation, is much faster than CP/M 2.2  running
on the same hardware.)
Additionally,  there  are  no built-in commands; all commands  are  loaded  as 
transient  programs from the disk. Hence, functions such as ERASE  and  RENAME
take  a bit longer to complete than their CP/M 2.2 counterparts, because  time
is  needed to load them. Most of the utilities, however, make up for  this  in
expanded  power and versatility, as we shall see. In any case, if a hard  disk
is used, the extra time is so small as to be unnoticeable.
From the command level, TurboDOS works similarly to CP/M in that commands  are 
typed  into  the  system, and the specified files are  executed  as  transient
programs.  If, however, your TurboDOS is part of a network, the  programs  and
files  that  you use are not necessarily located on your  local  computer  (in
fact,  your  local  computer may not even have any  storage  facilities  (disk
drives) at all!). This networking capability is one of the strongest  features
of TurboDOS, and we will examine that aspect of the system in detail later.

CP/M compatibility
------------------
TurboDOS  is  compatible with almost all CP/M programs. There are only  a  few 
CP/M  programs  that I have seen that do not work properly under  TurboDOS  --
mostly  public  domain  utilities that access some of the  more  obscure  BDOS
system  calls  (7, 8, 24, 28, 29, 31, and 100-107). However, TurboDOS  may  be
installed  with  an  optional CP/M-support module  that  will  simulate  these
missing functions.
Another  incompatibility  involves  CP/M's SUBMIT processing: I  am  aware  of 
several  commercial packages that make use of CP/M 2.2's special $$$.SUB  file
for batch-command processing, and that do not work under TurboDOS because  the
$$$.SUB  file  is  not processed. Another optional module  is  supplied,  that
modifies the TurboDOS batch function to emulate CP/M 2.2 more closely (but  at
the  expense  of slowing down DO file processing). The packages  mentioned  do
work properly under TurboDOS if this module is installed.

Multi-tasking the CPU
---------------------
Beginning  a more detailed study of TurboDOS, one of the first things that  we 
should  note  is  that  it can execute more than one  task  at  a  time;  this
capability  is usually referred to as *MULTI-TASKING*. (ROCHE> And  MP/M,  the
multi-user,  multi-tasking version of CP/M, had done it for at least  5  years
when this article was written...)
Integral to the concept of multi-tasking is the *RESIDENT PROCESS*. A resident 
process is a program that executes within the system; under TurboDOS, resident
processes  are  physically  attached to the operating  system,  but  logically
maintained  outside  of  it.  (That is to say:  they  perform  their  task  by
executing  operating system calls, in the same way that user programs do.  The
difference  is  that  user programs are transient in  nature,  while  resident
processes are always present.)
At  any  given time, a process may be in any of 3 states:  ready,  running  or 
suspended. A running process is one that is currently executing on the CPU;  a
ready  process  is  one that is waiting to execute (but  otherwise  ready).  A
suspended  process (sometimes referred to as "blocked") is a process  that  is
waiting  for some event to happen, or waiting for a period of time to  expire.
(TurboDOS provides a system call that allows a process to delay execution  for
a fixed time interval.)
The  suspended  state  is  perhaps the most significant.  A  process  that  is 
suspended  is  "stacked"  on  one  of  a  number  of  special  lists   (called
"semaphores"), and thus does not slow down the system by consuming CPU cycles.
At any given time, most processes will be in this suspended state. Hence, user
programs  that  are CPU-intensive (i.e., spend most of their  time  processing
data rather than waiting on I/O devices) run nearly as fast as they would on a
single-task system.
A  number  of "background" processes, present in each  CPU  running  TurboDOS, 
perform  such actions as monitoring the network, flushing disk  drive  buffers
periodically,   etc.   Additionally,  user-written  resident   processes   are
supported;  any  such  process is physically part of  the  TurboDOS  operating
system, and must be added to TurboDOS when the operating system is generated.

Banking the system
------------------
TurboDOS makes good use of bank-selectable memory. When more than one bank  is 
available,  most  of  TurboDOS resides in Bank  0,  while  transient  programs
execute  from Bank 1. (ROCHE> Exactly like CP/M Plus, the single-user  version
of  MP/M,  launched  one year before this article...)  Since  a  full-featured
network-master version of TurboDOS exceeds 24K in size, this frees up a lot of
program  area.  (ROCHE> That's why MP/M used banked memory, since MP/M  was  a
multi-user version of CP/M. So, each user wanted to have its own TPA, and most
CP/M program wanted as much RAM as possible... So, the only technical solution
was to provide "banks" of memory to each user.)
Banked  systems are often measured by the amount of *COMMON MEMORY* that  they 
consume.  Common memory is a section of memory (usually located physically  at
the top of the address space) that is always present within the CPU's  address
space,  regardless of which memory bank is switched in. Since the  Zilog  Z-80
CPU  is limited to 64K of total address space, it is obvious that the  common-
memory  requirement increases at the expense of the banked segments, and  thus
at  the  expense  of  memory available to  transient  programs  (in  the  case
of TurboDOS, at the expense of the allowable size of the operating system).
The  only portion of TurboDOS that *MUST* remain in common memory is  a  small 
inter-bank communications module, plus some portions of the  interrupt-service
routines.  (ROCHE>  Exactly like MP/M and CP/M Plus...) Thus,  the  amount  of
transient program area (TPA) available is 63K, even if banked memory  hardware
allows  bank  switching only on 4K or 16K boundaries.  (ROCHE>  Nothing  news,
here.  There is even a version of CP/M 2.2 for the Epson QX-10 having a  63.5K
TPA!...)
A  utility  program,  BANK.COM, allows the TurboDOS  operating  system  to  be 
switched between banked and non-banked mode. When non-banked mode is selected,
programs must share Bank 0 with the operating system.

Disk drives and files
---------------------
TurboDOS  supports file sizes of up to 32MB (ROCHE> Like CP/M Plus...)  and  a 
disk  drive  capacity of up to 134MB (ROCHE> CP/M Plus:  512MB...  Of  course,
compared to the 8MB of CP/M 2.2, this appeared to be an improvement...  Except
that  CP/M  Plus had been launched by Digital Research one  year  earlier!...)
Obviously, such large drives will have a lot of files, and will need a lot  of
seek  time  during  such operations as file opens,  directory  searches,  etc.
(ROCHE>  That's  why  directories are cached in CP/M Plus.) In  an  effort  to
decrease  this  time, TurboDOS supports a special directory  format  called  a
*HASHED  DIRECTORY*. (ROCHE> MP/M and CP/M Plus did not go this way.)  A  disk
directory  may, at any time, be restructured as a hashed  directory  (although
the opposite is not true) by running a program named FIXDIR. (ROCHE> CP/M Plus
has  something roughly similar, INITDIR.COM, which restructures the  directory
to  add date and time stamping of files.) Although hashing the directory  made
little difference on my small-capacity floppy disks, a hardware dealer I  have
worked  with  (who  sells TurboDOS on IMS systems) reports  a  dramatic  speed
increase  on his 40MB hard disk systems. (ROCHE> I am not surprised. Me,  from
experience, I only use INITDIR on floppy disks, since it consumes 1/4th of the
directory.  On the hard disk, I simply type the date and time at the  head  of
the files.)
Another technique used by TurboDOS to increase performance is disk  buffering. 
(ROCHE>  Called "disk caching" in CP/M Plus.) The system contains  a  built-in
buffer  manager;  moreover,  a  utility program  supplied  with  the  TurboDOS
operating  system, BUFFERS.COM, lets you change the size of each  disk  buffer
and  the number of buffers *WHILE THE SYSTEM IS RUNNING*. (ROCHE>  Under  CP/M
Plus,  this  is  done at system generation time, with GENCPM,  by  the  system
implementer.)  These buffers are taken "off the top" of the transient  program
area (TPA) (in unbanked versions of TurboDOS). If you are going to run  small,
disk-bound  programs, you can set up a large number of buffers,  and  increase
disk  access  speed.  Conversely, memory-hungry programs  (such  as  WordStar)
should  be  run with a small number of buffers, to provide as much  memory  as
possible.  (ROCHE> Well, with "banks" of memory, each user has its own 64K  of
memory...)
(I  should  point out that, while directory hashing  made  little  performance 
difference  in my system, increasing the number of disk buffers in the  system
increased  disk  throughput dramatically.) (ROCHE> That's  why,  after  tests,
Digital Research added disk caching to CP/M Plus.)
Systems  using banked memory may set up most of the unused part of Bank  0  as 
disk buffer space, and not affect program memory space at all.
A  fast file-load function supplies yet a 3rd performance increase:  generally 
used by the Command Interpreter (the portion of the TurboDOS operating  system
that processes user command lines), this module allows a section of memory  to
be  loaded  in  the fastest manner possible. It works  by  scanning  the  disk
allocation  map, detecting sequentially-allocated segments, and loading  these
segments  at the fastest transfer rate of the disk controller.  (ROCHE>  Under
CP/M  Plus,  the CCP is reloaded from a system Bank, so it is  just  a  memory
transfer,  and  using  the  "Multi-Sector Count", allowing up  to  16K  to  be
read/written in one shot...)

File attributes
---------------
TurboDOS  expands  on the file attributes employed by CP/M. (ROCHE>  ???)  The 
CP/M "SYSTEM" attribute is replaced by the "GLOBAL" attribute. Files  resident
in  User area 0 and tagged as GLOBAL are accessible from any user  area,  thus
circumventing the need to duplicate frequently-used files in each user area of
the  disk. (ROCHE> Under CP/M 2.2! But MP/M had been available for at least  5
years  when  this article was written, and CP/M Plus one  year!)  This  global
access  applies  equally to files that are accessed  by  application  programs
(such as WordStar's overlay files), or chained BASIC programs.
ARCHIVE  is another (ROCHE> CP/M...) file attribute that saves time and  space 
when  backing up files. New files are created with the ARCHIVE attribute  off.
When a file is backed up using the COPY program, the ARCHIVE bit is set;  when
a  file  is changed (extended, or written to with a  random  write),  TurboDOS
resets  the ARCHIVE attribute. COPY can be instructed not to copy  files  that
have  been archived and have not been modified since. This process  is  called
"incremental backup". (ROCHE> And is provided by CP/M Plus' PIP...)
The final "new" file attribute is FIFO (ROCHE> This is specific to TurboDOS.); 
it  defines  a  special type of file accessed using  a  "first-in,  first-out"
(FIFO)  technique. FIFO files are accessed from programs like any other  file,
except that sequentially-written records are appended at the end of the  file,
and  sequential reads are taken from the beginning. Moreover, a FIFO file  may
be  declared as an in-memory file (accessed much faster, since it really  does
not  exist  on  disk at all; this limits the size of  the  file,  however,  to
available  memory.) (ROCHE> Most CP/M Plus implementations have a RAMdisk,  so
the  limit  is 32MB, in this case.) This whole scheme provides  a  very  handy
technique for passing data between background processes, and between  programs
executing on different processors. Since a FIFO file can be made to  "suspend"
(deactivate)  a process reading the FIFO file when it is empty, it is easy  to
visualize a background process that only "comes alive" when a program posts  a
record to a file monitored by the background process.
Two special utilities are provided to work with FIFO files: SEND and  RECEIVE. 
SEND  is invoked with a FIFO filename and a text string to place in the  FIFO.
RECEIVE  takes  only a FIFO filename; it reads one record from the  FIFO  file
(thus deleting the record) and displays it. Together, these 2 utilities may be
used to form a rudimentary mail facility between users. (ROCHE> Better to  use
CP/NET, which has been available for years...)

File interlocks
---------------
TurboDOS  features  file  and record lockouts, since  it  supports  background 
processes  that may access files at the same time as foreground processes.  In
fact,  network support for shared files requires at least one such process  in
each  file system offering file resources to the network. Interlocks are  used
to coordinate file accessing by multiple simultaneous processes and necessary,
for example, when an update program must replace an inventory item's wholesale
and  retail price fields. With no coordination, a querying program might  read
the  record sometime after the update program has begun writing new data,  but
before  the write is complete. Hence, the querying program reads  inconsistent
data.  With record locking capability, however, the update program can  obtain
exclusive  access to the record, "locking out" any other process that  may  be
attempting to access the same record.
File lockouts under TurboDOS may be configured for strict MP/M  compatibility; 
thus,  existing programs written for MP/M's record-locking techniques  (ROCHE>
Like Digital Research's Access Manager...) will work correctly under  TurboDOS
(ROCHE>  Fortunately,  since TurboDOS is a multi-processor  version  of  MP/M.
Hence the need to be compatible with its parent. It is the classical  "chicken
and  egg" problem.) Alternatively, a set of compatibility flags  is  available
(during system generation) to change certain rules, providing a number of  new
modes  not  previously available under MP/M. (ROCHE> Modes  which  disappeared
with TurboDOS, while CP/M is still alive, 30 years later...)

Networking
----------
TurboDOS  provides a plethora of features, facilities, and utilities, but,  to 
me,  the  single most outstanding one is its built-in  networking  capability.
(ROCHE>  Never  heard of MP/M, launched 5 years earlier by  Digital  Research?
Never  seen  the advertisements for CP/NET in computer magazines? Why  do  you
think that TurboDOS is compatible with MP/M? For pleasure?) Multiple computers
running  TurboDOS  may be interconnected and share such  relatively  expensive
peripherals as disk drives, printers, plotters, etc.
In  this  discussion,  I  will use  the  terms  "node","CPU"  and  "processor" 
interchangeably.  (ROCHE>  Not a good thing. It shows that the  concept  of  a
network  is  still  a novelty for you. You should have read  the  MP/M  guides
published  several years earlier...) Generally, a node can be thought of as  a
workstation containing a CPU, memory, a network interface, and (optionally)  a
terminal, disk drives, and printers. A node is attached to the network via one
or  more  connection  points. It usually has some peripherals,  but  does  not
*REQUIRE* any.
Two  other terms will be used to refer to processors on the network:  *MASTER* 
(or  *SERVER*)  and *SLAVE* (or *SATELLITE*). Generally,  a  server  processor
provides  resources  to the network (such as disk  storage  and/or  printers),
while  satellites  do not. If satellites have local disks, these can  also  be
available to the network if the satellite contains the network service  module
(NETSVC).
First,  it is important to note that TurboDOS does not support multiple  users 
on  a single CPU. (ROCHE> Yes, since TurboDOS is a multi-processor version  of
MP/M.)  The  philosophy,  here,  is that,  since  processors  and  memory  are
relatively  cheap, why attempt to split up a hundred dollars or so in CPU  and
memory  hardware  when each user can have his own CPU and memory,  and  suffer
none of the processing delays that are so severe in such operating systems  as
MP/M  and  OASIS?  (ROCHE>  Good  question.  By  the  way,  why  did  TurboDOS
disappear,  while  MP/M  and OASIS are still running, 30  years  later,  under
emulators on the Internet?)
This concept works, and works well. (ROCHE> You should tell Intel, which never 
managed  to  make  its 8086 works this way. I  cannot  resist  mentioning  the
following, taken from "Chapter 1: Introduction" of the "Intel 8088/8086 Family
User's Manual", October 1979 (Yes, 5 years *EARLIER*!), page 1-1:
        "The 8086 family architecture is characterized by 2 major principles:
     1. System functions are distributed among specialized components.
     2. Multi-processing capabilities are inherent in the hardware.
     3. A  hierarchical bus organization provides for the complex  data  flows 
        required by high-performance systems without burdening simpler systems
        with unneeded capabilities."
This was written in 1979, for a CPU running at 4-MHZ. Today, *45 YEARS LATER*, 
Intel  has  not  been  able to produce even  a  single  computer  using  those
"principles"...  So, why did Intel, the richest microcomputer company  in  the
world, never managed to do it, while Software 2000, Inc. (e.g., Mike Busch and
Ron Raikes), did it? Was the Zilog Z-80 CPU so much better than the Intel 8088
of the IBM Clown?)
Each  user has full use of his own CPU; delays are generally  noticeable  only 
when shared disk drives are simultaneously accessed by multiple network users.
(Even   these  delays  can  be  abated  by  using  faster  drives  with   more
sophisticated seek mechanisms -- a cost that can be justified more easily with
TurboDOS,  since these expensive peripherals are more easily shared among  all
the users.)
TurboDOS  supports just about any kind of local network, ranging  from  simple 
implementation involving 2 computers (liked through,say, an RS-232-C channel),
to sophisticated structures with many processors sharing resources over  high-
speed  data  channels. The actual network structure is defined by  the  system
designer,  and  implemented  in  the network  driver  modules.  These  modules
constitute  what is known in network vernacular as the "physical layer";  they
control the hardware that interconnects the computers.
TurboDOS  defines  the network as being composed of *CIRCUITS* (which  can  be 
better  visualized as a cluster of nodes). A circuit may be composed of up  to
255  nodes. Further, there may be up to 255 interconnected circuits in a  full
TurboDOS system, allowing for a total of up to 65,535 total workstations in  a
TurboDOS network.
The unit of information exchange among network nodes in the *NETWORK MESSAGE*. 
This  message  is  a packet of data, containing source  and  destination  node
addresses,  byte length and other overhead, as well as the actual data  itself
(which may be a disk record, a file control block (FCB), or perhaps a block of
characters to be sent to a printer).
A  "forwarding" table may be specified in each node, allowing messages  to  be 
sent between network nodes that are not, themselves, directly connected.
One of the most common techniques that I have seen for implementing a  network 
with  TurboDOS  involves  S-100 Bus  single-board  computers.  Generally,  one
"master"  CPU is provided (to control access to disk drives, printers,  etc.),
and  one  or more "slave" processors are placed on the same S-100 Bus  as  the
master  (often these slaves have no local storage of their own  but,  instead,
rely  on  the master to provide storage. The user will see the  master's  disk
drives  as  if  they were connected to his  own  CPU.)  Typically,  high-speed
parallel-port  channels  are provided to allow network transfers  at  near-bus
speeds.  One  such system will, generally, be a circuit  in  itself;  multiple
systems may, then, be interconnected, forming a multiple-circuit network.
        Figure 1. A typical network
An  example should help tie this all together. In Figure 1, we have 2  network 
circuits  that  are  relatively independent. Each circuit  has  a  master  and
several  slaves,  as  well  as disk  drives  and  a  printer.  Master-to-slave
communication   is  through  a  high-speed  parallel   bus.   Master-to-master
communication is through a Local Area Network (LAN) interface and cable.  Note
that  slave  nodes do not have any local disk storage but,  instead,  use  the
master's  disk  drive (this will likely be a large-capacity hard  disk  drive,
segmented  into  several smaller drives, and apportioned  among  the  slaves).
Circuit 1 consists of 4 nodes: the master and the 3 slaves. Similarly, Circuit
2  consists of 5 nodes: the master and 4 slaves, 3 of which are connected  via
modems to dial-in phone lines.
Each  of the 2 masters are actually "known" on 2 circuits (their own, and  the 
other  master's)  and,  thus, have 2 circuit  drivers  within  their  TurboDOS
operating systems (and 2 network addresses, as a matter of fact).
This  system  might be part of an overall data-gathering  system  where  field 
agents phone in information to the nodes connected to modems (Nodes 2-2,  2-3,
and  2-4);  the  high-speed dot-matrix printer might be  used  to  maintain  a
"running  log"  of  these transactions. This raw  information  is  sorted  and
formatted  by programs entered by an operator at the terminal located at  Node
2-1 (the only "on-site" terminal in Circuit 2) and made available periodically
to Circuit 1 through the RS-232-C link between circuits, where service  clerks
work  at  terminals,  filling these orders. The  finished  orders  are,  then,
printed  up in the form of invoices on the letter-quality printer  located  on
Circuit 1.
Thus, most information flow between the 2 circuits will take place between the 
2  masters  (perhaps under the control of a resident process in each).  It  is
possible,  however,  to  allow  communications  between  slaves  on  different
circuits, using forwarding tables contained in each slave's operating  system.
This  might  be necessary, for example, to allow field salesmen  calling  into
Circuit  2 to make an inquiry about an order to a clerk working in Circuit  1.
Proper setup of the forwarding tables will allow this (in fact, the Turbo-Plus
enhancement  package  provides  special system functions  and  utilities  that
simplify this kind of communication).

Reconfiguring the network
-------------------------
Although  the physical network is defined in network driver modules,  TurboDOS 
provides  a lot of latitude in reallocating resources, using  the  previously-
mentioned  GEN  program.  Although  these  resources  are  not   *DYNAMICALLY*
changeable  (that  is  to  say: you cannot make  changes  while  the  TurboDOS
operating  system is running), it is a relatively simple matter to  patch  the
tables, and regenerate the system.
        Figure 2. Drive allocation example
An  example  should  help to clarify this. Referring to Figure  2,  we  see  2 
processors, named P1 and P2, each with local drives A and B. From the diagram,
you  can  see  that  each  processor has a drive  C,  which  is  the  opposite
processor's B drive. (These assignments are made in a table contained in  each
node's  copy of TurboDOS; this table is called the Disk Assignment Table,  and
is usually set up in a PAR file.)
        Figure 3. Allocation after regeneration
Now,  let  us consider changing that. Let us say that we want to  delete  P2's 
ability  to  reference P1's B drive, and increase P1's  available  storage  to
include P2's drive as P1's D. We simply make a few changes in each node's  PAR
file,  regenerate  each TurboDOS operating system, and we have  the  structure
shown  in Figure 3. This concept applies to other resources, such as  printers
and print queues.

Printers and spooling
---------------------
Flexibility is apparent throughout the design of TurboDOS, but nowhere is this 
more evident that in the facilities available to produce printed output.
Each  processor  may have up to 16 printers defined in its  tables,  and  each 
printer  may  be accessed either locally or over the network.  Since  TurboDOS
provides a CP/M-like environment, only one of these printers may be  "current"
at  any  one  time for any one user. A utility program,  PRINT.COM,  lets  you
change the routing of your logical printer output, using syntax such as:
        PRINT PRINTER=E   (or PRINT P=E)
which selects Printer E in your local printer table. In addition, this utility 
allows  you  to  select  a  much more  flexible  form  of  printing:  *SPOOLED
PRINTING*.  (ROCHE>  Ever  heard of the "DESPOOL  (background  print  utility)
Operator's Guide", Digital Research, 1979? No? It was frequently found in CP/M
Version 1.4 and Version 2.2 advertisements...)
Spooled  printing reroutes your printer output: instead of going  directly  to 
the  printer, the output is routed to one of 16 *PRINT QUEUES*. A print  queue
is  implemented  as a set of disk files, using filenames of the  form:  PRINT-
x.nnn,  where  x  is  the queue name (A-P, corresponding to  each  of  the  16
possible queues) and nnn is a 3-digit number used to separate different  print
jobs within a queue.
Another  utility, PRINTER.COM, controls despooling of these print queues.  Any 
physical  printer may be assigned to any of the queues with a command such  as
PRINTER  E  QUEUE=C  (to  assign Printer E to the C  Queue).  A  3rd  utility,
QUEUE.COM, allows files to be manually entered into any of the various queues.
Let us tie this all together with an example: suppose your TurboDOS  operating 
system has 2 printers: Printer 1 is a fast dot-matrix printer, while Printer 2
is a slower letter-quality model (remember, either printer may be attached  to
our local network node, or to a remote node -- it does not matter *WHERE*  the
printer to be used is actually located). Assume that we have to print  several
program listings, a couple of letters, and a long report.
First,  we  will tackle the letters, since they will likely  use  single-sheet 
paper, requiring a good deal of manual intervention. The command:
        PRINT PRINTER=B
gives  up  direct access to the letter-quality printer (we  must  be  careful, 
here, to coordinate access with any other network users, since direct printing
by  more  than one user will result in interspersed characters in  the  actual
printout). Now, we fire up the word-processor program, and print each  letter.
Not  so  different  from  CP/M,  so far (except  for  the  ability  to  switch
printers).
Now,  let  us tackle the program listings. We can begin to make use  of  print 
queues, here, by entering the following commands:
        PRINTER A QUEUE=A
        PRINTER B QUEUE=B
Now,  we have established Queue A as the queue associated with the  dot-matrix 
printer,  and  Queue B with the letter-quality printer. Now, we are  ready  to
print some listings. The command:
        PRINT QUEUE=A
routes our printer output to Queue A; now, we can run our language  translator 
and  begin producing the program listings. Since our output is now going to  a
relatively  high-speed disk file (as opposed to a slow printer), the  language
translator finishes long before the first printout is complete.
In  fact, the first output file does not begin printing until  its  associated 
queue  file  is  closed (either by the  language  translator  terminating  and
returning to the TurboDOS operating system, or by a special end-print  control
sequence  from  the  console).  While  the first  file  is  printing  (in  the
background), we can run our language translator once more, to generate the 2nd
and 3rd listings. We might also decide to queue some existing disk files:
        QUEUE MYPROG.PRN ;Q=A
        QUEUE MANUAL.DOC ;Q=A
        QUEUE *.UPC ;DYQ=A
(Note  that the Q=A options specify which queue to send the file to.  The  3rd 
example illustrates queuing multiple files by specifying a wildcard  filename.
Additionally,  the ;DY portion specifies options. In this case, D causes  each
file to be Deleted after printing, and Y causes the QUEUE program to displa(Y)
the name of each file to be queued, allowing the operator to select  specified
files.)
Now, we are ready to print the letter-quality documents. The command:
        PRINT QUEUE=B
switches  our printer output to the queue we have previously  associated  with 
the  letter-quality printer. (It does not matter that the files that  we  have
already  queued are still printing; they are now destined for  the  dot-matrix
printer, and changing the destination of our printer output no longer has  any
effect  on the previously-queued files.) Now, we can use  our  word-processor,
and  send  the letters and the report to the B Queue, similar to the  way  our
language translator output went to the A Queue.
At  this point, we have 2 printers running a number of print jobs entirely  in 
the background, and our console is free for new work. In addition, any  number
of network users can simultaneously queue files on the same printers,  without
intermixing printer output.
To  conclude  the discussion, I must mention that there are  many  more  print 
options  and  features  than those mentioned here; to  review  each  of  them,
however, is beyond the scope of this article.

Batch files
-----------
TurboDOS  allows  the  execution of batch-command files in a  way  similar  to 
CP/M's  SUBMIT facility, but with enhanced capabilities. The TurboDOS  version
of SUBMIT is named DO.COM; specifying DO with a filename will cause the system
to begin reading its input from the file.
DO  files  may  be  nested (ROCHE> Like SUBMIT under  CP/M  Plus...),  to  any 
reasonable depth (ROCHE> 128 levels under CP/M Plus.), merely by placing a  DO
statement within a DO file. After the subservient DO file is executed, control
returns to the superior file at the point where processing left off.
Command-line parameters may be specified in a manner similar to CP/M's SUBMIT. 
However, the parameters are referred to, in the DO file, by enclosing them  in
curly  brackets,  rather  than the dollar sign notation  of  CP/M.  Also,  the
parameters  in the DO command line may be made to contain embedded  spaces  by
enclosing them in quotes.
A handy option within the DO file allows default parameters to be filled in by 
adding  a  comma and a default value within the curly brackets.  Consider  the
following line within a DO file:
        L80 {1} {2,SYSLIB}
If  this  DO file is invoked only with Argument 1, then the  value  SYSLIB  is 
automatically supplied as the 2nd Argument.
Another  important characteristic of DO files is that they can  provide  input 
for  more than just system commands. Programs that require  interactive  input
will  receive that input from the DO file (thus, DO combines the  features  of
both  the  CP/M 2.2 programs SUBMIT and XSUB). (ROCHE> CP/M Plus  has  a  much
better and more powerful SUBMIT utility.)
DO  is somewhat faster than CP/M 2.2's SUBMIT when no command-line  parameters 
are  specified. Since there is no need to rewrite the DO file with  parameters
filled in, DO simply informs TurboDOS to begin executing the file (SUBMIT must
always write a $$$.SUB file).

Utilities
---------
TurboDOS  is  supplied with a large number of utilities, some  of  which  have 
already been described. The major remaining utilities are as follows:
AUTOLOAD  allows a user to set up command lines that automatically execute  on 
either  cold  or warm boot. COLDSTRT.AUT, the cold boot  file,  executes  only
once, when the TurboDOS operating system is first started up. WARMSTRT.AUT  is
executed at the completion of each transient program.
BACKUP  performs  a  fast disk copy. It is usable only  when  the  source  and 
destination diskettes are of the same format.
BATCH  is useful in network environments, and uses a special FIFO file  (named 
BATCH.DO)  to send a command line to a remote processor for execution.  Remote
processors  that use this facility (i.e., those which are to be available  for
batch  jobs) must execute the command DO BATCH via an AUToload file at  system
startup.
BOOT allows the system tracks of a diskette to be copied to another  diskette. 
Note  that TurboDOS allows 100% of the floppy disk space to be used  for  file
storage; however, in some implementations, the hardware demands  cold-starting
from reserved system tracks.
CHANGE  is used to safely allow removable disks (i.e., floppies  or  removable 
cartridges)  to be changed in a network environment. CHANGE.COM takes  one  or
more  drive  letters  as arguments, and aborts with an error  message  if  any
process has open files on the requested drive. If the requested drive is free,
CHANGE locks the drive until the user indicates that the disk has been changed
(any  user trying to access the drive while it is locked will receive  a  disk
error message).
COPY,  which was previously mentioned, allows files to be copied between  disk 
drives.  A  large  number  of command-line  options  make  this  an  extremely
versatile file copy and archiving program.
DELETE performs the same functions as CP/M's ERA. A command-line option causes 
DELETE  to  present each filename to be erased, and requests the  operator  to
confirm  before  deleting. (ROCHE> "The CONFIRM option informs the  CP/M  Plus
operating  system  to prompt (the user) for verification before  erasing  each
file  that matches the filespec." CP/M Plus User's Guide, Section  5:  Command
Summary, "The ERASE command".)
DIR prints a sorted directory on the screen (or, optionally, on the  printer). 
The  display  contains  a lot of the information  available  under  CP/M  only
through the CP/M 2.2 STAT command (such as filesizes and disk free space).  It
also  includes  time of day and combined size of files displayed, as  well  as
user-number specification.
DRIVE  displays information about the disk format (similar to CP/M 2.2's  STAT 
DSK:).
FIFO is used to create and delete the previously-described FIFO files.
FIXDIR  is a handy utility that reorganizes a disk directory. This  is  useful 
when  frequent file creations and deletions occur (especially on hard  disks).
FIXDIR  compacts  the directory, eliminates "holes", and makes  file  searches
faster. Also used to convert a directory to the hashed format, or  vice-versa.
(ROCHE>  ???  Ron  wrote "the opposite is not true" in the  "Disk  drives  and
files" paragraph...)
FORMAT  allows  a  diskette to be formatted in single density or  in  CP/M  or 
TurboDOS double-density format. It automatically calls VERIFY after formatting
is complete. (ROCHE> VERIFY marks all the "bad spots" of a disk. That's why it
is used after formatting the disk.)
MONITOR  is  the  TurboDOS replacement for CP/M 2.2's  DDT  program.  Intended 
mainly  for program patching, etc., it contains a save function as well  as  a
load  function  (this  is necessary because TurboDOS  lacks  CP/M  2.2's  SAVE
command). A number of other commands not found in DDT are present in  MONITOR.
(ROCHE> ASM and DDT were provided with CP/M 2.2. MAC and SID were professional
versions that were available separately, at a cost.)
Noticeably  lacking,  however,  are DDT's opcode  assembler  and  disassembler 
(ROCHE>  The A and L commands.), and single-step trace functions  (ROCHE>  The
TRACE  utility  of  SID.).  For program  debugging,  DEBUG  (Phoenix  Software
Associates (ROCHE> Software 2000, Inc., used their PASM to assemble TurboDOS.)
or  ZSID (Digital Research, Inc.) both work well under TurboDOS. (ROCHE>  This
is no surprise...)
RENAME  is  used  to  change the names of files. This  utility  has  a  syntax 
backward from that of CP/M. (Most people who have worked with other  operating
systems  claim that it is CP/M that has it backward. I am inclined to  agree.)
(ROCHE>  You (and them) are wrong. Demonstration: How work the opcodes of  the
(real, physical) processor? You write MNEMONIC DESTINATION,SOURCE... There  is
no  exception  to  this  rule!  So,  is  the  CPU  behaving  crazily?  Another
demonstration:  When you send dozens of different files to the same place.  If
the  destination  is at the beginning of the command line, you just  edit  the
source  specification, conveniently located at the end of the command  line...
which can be reached with one single control char!) This utility has the handy
option  of renaming by using wildcards. For example, the command RENAME  *.BAK
*.ASM renames all BAK files to ASM (a command-line option allows you to verify
any  rename  that  would  cause  an existing file  of  the  same  name  to  be
overwritten). (ROCHE> "If the file given by NEWNAME is already present in  the
directory,  RENAME displays the following message on the screen: Not  renamed:
FILENAME.TYP  file  already  exists, delete (Y/N)?" CP/M  Plus  User's  Guide,
Section 5: Command Summary, "The RENAME command".)
SERVER  has been renamed by North Star. The standard TurboDOS  distributed  by 
most vendors refers to this utility as MASTER.COM. In any case, its purpose is
to allow a slave console to act as a terminal to the master, a function useful
for running programs in the master processor.
SET is used to set and clear file attributes; SHOW displays them. (ROCHE> Just 
like MP/M and CP/M Plus...)
TYPE displays ASCII files on the console or (optionally) on the printer.
USER allows privileged logins to change the currently-logged user number. This 
is one command that I wish had been built into the TurboDOS operating  system.
(ROCHE>  Ron  ported  TurboDOS  Release 1.3. As far as  I  know,  the  Command
Interpreter of TurboDOS Release 1.4 recognizes the small form of changing user
numbers (1:), without the need of running the transient program USER.COM.)
VERIFY scans a disk for bad blocks, and locks them out.

Program interface
-----------------
As you would expect, TurboDOS provides a good many additional system calls for 
the  programmer.  (ROCHE> Well, TurboDOS being a  multi-processor  version  of
MP/M,  most of them have been existing under MP/M for about 5  years...)  Non-
programmers may want to skip this section.
First of all, CP/M-compatible system calls are vectored through location 5 (as 
you would expect); versions of TurboDOS prior to Release 1.3 included extended
system  calls  vectored  in the same way. (ROCHE>  Like  MP/M...)  Recent  re-
releases  of CP/M and MP/M, however, have caused some conflicts with  TurboDOS
extended functions. (ROCHE> ???)
To  correct  this problem, TurboDOS Release 1.3 moved  all  TurboDOS  extended 
system calls to location 80 (0050H). CP/M-compatible (ROCHE> Which version  of
CP/M?  2 or 3?) system calls are now referred to as C-functions, and  TurboDOS
extended system calls as T-functions.
Resident  processes must often make system calls; they cannot call location  5 
or  50H.  (ROCHE>  ??? Why?) Two special entry points are  provided  for  this
purpose,  and must be called symbolically. They are: OCNTRY (C-functions)  and
OTNTRY (T-functions). (ROCHE> So, O means Operating system, C means CP/M,  and
T  means TurboDOS.) The calling conventions are, otherwise, strictly the  same
as  they would be for a transient program. For example, to set the DMA  buffer
address, you would use the following code:
        MVI     C,SetDMA
        LXI     D,Buffer
        CALL    OCNTRY#
A  wide range of T-functions is provided (see Table 1) for such varied  things 
as  creating  and deleting resident processes,  allocating  and  de-allocating
memory,  inter-process communications, date/time functions, extended file  and
disk  drive  requests, and more. A few of the more interesting  functions  are
described here.
Communications  channel  support  (T-functions 34-40): A  full  complement  of 
functions  allows  complete  access  to  the  optional  communications  system
(generally  implemented  in  the  form  of  RS-232-C  drivers  by  the  system
implementer).  These functions allow such actions as  send/receive  character,
read  status, get/set baud rate, and get/set modem controls. They  also  allow
the  creation  of system-independent communications software for  TurboDOS  --
something  that  is impossible with CP/M. (ROCHE> ??? And CP/NET, what  is  it
doing? It is not tied to any protocol or hardware!)
Delay  process (T-function 2) circumvents the need for software timing  loops. 
This function allows a program (or resident process) to delay by increments as
fine as the system "tick" time (generally 1/60th of a second). (ROCHE> In  the
USA. 1/50th of a second in Europe.)
Return serial number (T-function 12) returns the complete serial number of the 
TurboDOS operating system. (This function can be used to discourage piracy  by
keying the purchaser's TurboDOS operating system to your software package.)
Load file (T-function 15) uses the previously-described Load Optimizer to load 
code segments into the program area. Handy for loading overlays.
Activate DO file (T-function 16) allows a program to begin a batch  processing 
job.  (Compared  to  writing  the old backward $$$.SUB  file  with  CP/M  2.2,
starting batch jobs under TurboDOS is simple with this system call.)
Send  Command  Line (T-function 18) allows creation of  applications  such  as 
menus from assembly language programs. It does not work quite the same way  as
MP/M's similarly-named function (ROCHE> This must be a coincidence?), however,
in that the passed Command Line is not executed immediately (as it is in MP/M)
but  is,  instead,  deferred until the calling  program  terminates.  You  may
"stack" command lines by calling this function more than one.
Printer  control  functions  T-functions 27-30)  allow  complete  programmatic 
access of the printer spooling and despooling mechanisms. Function 28  (signal
end-of-print)  is  especially  handy, since it allows a queued  print  job  to
actually begin printing.
User-defined  function  (T-function 41) belongs to the user, for  any  purpose 
that  he desires to write code for. It can take full advantage of the  network
for  communications between processors. To implement this function,  you  must
write  your  function  in  a module that defines  the  public  symbol  USRFCN;
register-passing  conventions are described in the "TurboDOS Release  1.4  Z80
Programmer's Guide". Once the function is in place, your applications  program
may communicate with it simply by calling T-function 41.

Documentation
-------------
No  software  evaluation  can be considered complete without  a  look  at  the 
documentation  supplied with the package. In the case of TurboDOS, I  have  to
rate  the documentation effort outstanding! (ROCHE>  Interesting.  Apparently,
not everyone thinks like you. "Many features of TurboDOS go underutilized  (or
totally  ignored) due to the difficulty in deciphering the manuals." and  "the
manuals  are  indescridably bad." This opinion was published in  nothing  less
than  the  "ACM SIGSMALL Newsletter"! Reference: Vol.10,  No.3,  August  1984,
p.24, "The TurboDOS Operating System", Keith Bierman.)
Three  manuals are provided with the Z80 TurboDOS operating system:  a  User's 
Guide, a Z80 Programmer's Guide, and the Z80 System Implementer's Guide.  Each
includes  a  table  of contents; the System Implementer's Guide  is  the  only
manual lacking an index.
The  User's  Guide  begins  with basics, and  takes  the  reader  through  the 
gradually  more  complex subjects of files, disks, printing,  and  processing,
concluding  with  a  summary of the entire set of 35-odd  utilities  that  are
provided with the TurboDOS operating system. The language is "gentle" and very
readable.
The  Programmer's  and Implementer's Guides are understandably  more  complex, 
since they deal with subjects that are technical in nature. They are complete,
and  provide  sufficient information to enable the programmer to do  his  job.
(ROCHE> Again, Ron implemented TurboDOS Release 1.3. Release 1.4 adds the 8086
programmer's Guide and the 8086 Implementer's Guide. TurboDOS thrived when  it
was  written in Z-80 assembly language for S-100 Bus systems, and  disappeared
when it was rewritten in the C language for the IBM PC...)

Conclusion
----------
As  you see, TurboDOS provides powerful facilities for both the user  and  the 
programmer.  After more than 3 years on the market (it was first  released  in
April  1981)  (ROCHE>  So, MP/M had already been existing  for  2  years  when
TurboDOS appeared... Guess which is the chicken? Which is the egg?),  TurboDOS
is  still being continually refined. (By the time you read this,  Release  1.4
should  be  available.  This new release will  feature  full  Intel  8088/8086
compatibility  at both the slave *AND* the master level, as well as a  PC  DOS
emulator.)  (ROCHE> Emulating MS-DOS Version 1 under TurboDOS on a  S-100  Bus
system? Vade Retro, Satanas!)
I  have recently obtained a Mega Z-80 system, featuring  512K  bank-selectable 
RAM  and  27 *MEGABYTE* Winchester hard disk drive, for which I plan to  do  a
full  master-to-master implementation of TurboDOS (complete with RAMdisk).  As
"Microsystems"  magazine  expands its TurboDOS coverage, I hope to  provide  a
follow-up article on this project, as well as a review of Turbo-Plus, and  (if
there  is  sufficient  interest) a tutorial  on  TurboDOS  implementation  and
programming  technique. (ROCHE> This article was the last one written  by  Ron
Fowler...)

EOF
TDOS7.WS4
---------
- "Printer Networking with TurboDOS"
  Tedd Kurts, MuSYS Corporation
  "Microsystems", Vol.5, No.10, October 1984, p.106
(Retyped by Emmanuel ROCHE.)

In working with TurboDOS users on a daily basis, I have observed many of  them
attempting  printer configurations, only to find their  efforts  unsuccessful.
The  problem  appears  to stem from a lack of  understanding  of  the  printer
networking concepts involved. This article will outline the steps necessary to
configure a TurboDOS operating system with one or more printers.
There  is  a  radical difference between CP/M or MP/M  operating  systems  and 
TurboDOS.  No printer drivers (or only skeletal examples) are supplied with  a
CP/M  or MP/M as purchased from Digital Research,Inc. The user must write  his
own drivers, and integrate them into the I/O section of the operating  system;
he  must,  therefore,  be  a  competent  assembly  language  programmer,   and
understand  the  workings  of the CP/M BIOS or MP/M XIOS. (ROCHE>  I  had  the
curiosity  of searching for "Printer" in the Index of "The  CP/M  Programmer's
Handbook",  by  Andy Johnson-Laird, Osborne/McGraw-Hill, 1983, the  only  book
ever  written  on  the subject of writing a BIOS for CP/M. There  are  only  3
entries:  "Printer echo", "Printer errors", and "Printer timeout error".  This
is  because  printers are, fundamentally, outside  the  microcomputer  running
CP/M.  Hence,  only CP/M programs dealing with printers (like  WordStar)  have
(their own) printer drivers.)
TurboDOS,  on the other hand, comes with several printer drivers  for  various 
methods  of handshaking; these are assembled in Microsoft RELocatable  format,
and need only be linked to the body of the TurboDOS operating system by  means
of  the  GEN  (system generation) utility (also supplied).  For  fine  tuning,
PARameter  files can be created or modified with a simple text  editor.  Thus,
the  person  who  "generates" a bootable TurboDOS operating  system  from  the
modules  supplied  in  Microsoft REL form need not  be  an  assembly  language
programmer  (ROCHE>  ??? How many end users are familiar  with  Microsoft  REL
files  and  linkers?), though he must understand what  functions  the  printer
drivers and PARameter files perform, and how to modify them to suit his needs.
(ROCHE>  Isn't  it  the  job of the  "System  Implementer"  to  port/implement
TurboDOS on a particular hardware, so that the user can use the software?)
The procedure falls into the following steps:
      - Determine the handshaking required by your printer(s), and select  the 
        appropriate driver module(s).
      - Determine  the baud rate to be used, and create a corresponding  patch 
        in a PARameter file.
      - Assign the printer to a spooling queue by a patch in a PARameter file.
Exampes will be given, showing exactly how to perform theses steps.

1) Printer handshaking and cabling
----------------------------------
To  determine  what handshaking protocol should be  used  (and,  consequently, 
which printer driver module to select), consult your printer manual. The  most
commonly-used  methods are hardware handshake (sometimes called CTS (Clear  To
Send)  protocol);  XON/XOFF protocol; ETX/ACK protocol; or  a  combination  of
hardware handshake and one of the other two (ROCHE> Software handshakes.). All
of  these  can  be  used with serial printers;  a  special  form  of  hardware
handshake,  standardized  by  Centronics,  is  generally  used  for   parallel
printers.
When  hardware  handshaking  is used on an RS-232-C  serial  channel,  only  3 
conductors  are required: the data line (from computer to printer); a  control
signal line; and the signal ground line. The printer accepts characters  until
its  buffer  is  about 3/4 full; it then turns off  the  control  signal.  The
computer  must  be  able to detect this signal, and stop  sending,  until  the
printer  turns  the  signal on again, when the buffer  is  nearly  empty.  The
control signal chosen is usually DTR (Data Terminal Ready, pin 20 of the  1969
RS-232-C  standard)  or RTS (Request To Send, pin 4), but  some  printers  use
other signals. This is the industry-preferred protocol, for several reasons:
     1) only one data channel is needed;
     2) it  is a hardware handshake and, therefore, the driver is  simple  and 
        takes less CPU time;
     3) a  cable  disconnection during printing does not cause loss  of  data. 
        When the cable is reconnected, printing picks up where it left off.
When  XON/XOFF  protocol  is  used, 2 data channels  are  required:  one  from 
computer  to printer, the other from printer to computer. The printer  accepts
characters  until  its  buffer is nearly full, and then sends  the  XOFF  code
(Transmit OFF, or "pause transmission", ASCII DC3, Control-S, 13H) to tell the
computer to stop sending; when the printer is ready to accept more characters,
it  sends  the  XON code (Transmit ON, or "resume  transmission",  ASCII  DC1,
Control-Q, 11H).
When  ETX/ACK  protocol is used, 2 channels are again required.  The  computer 
sends  a fixed number of characters to the printer (amounting to about 75%  of
the  buffer capacity) followed by an ASCII ETX (End-of-TeXt,  Control-C,  03H)
character; it then waits until it receives an ASCII ACK (ACKnowledge, Control-
F, 06H) character from the printer.
Both XON/XOFF and ETX/ACK can lose data in the event of a cable disconnection. 
For this reason, they are sometimes combined with a hardware handshake.
If the printer can use more than one of these protocols, selection is done  by 
setting switches or placing jumpers on the circuit board.
Care must be taken to see that the cable connects the correct pins of the  RS-
232-C connectors at each end. Devices designated DTE (Data Terminal Equipment)
send  data  on  pin 2 and receive it on pin 3; devices  designated  DCE  (Data
Communications Equipment) receive on pin 2 and send on pin 3. Thus, the  cable
connecting  a  DTE  device  to a DCE  device  has  corresponding  pin  numbers
connected.  If, however, both devices are DTE or both are DCE, then the  cable
connecting them must have pins 2 and 3 at one end connected to pins 3 and 2 at
the other. Also, if the control signal appears on (say) pin 14 at the  printer
but is required on pin 20 at the computer end, the cable must have this cross-
connection also.

Printer drivers
---------------
For  each  of  the protocols described above,  TurboDOS  has  a  corresponding 
driver.  The  most  common  ones  are  designated  LSTCTS.REL  (CTS   hardware
handshaking);  LSTXON.REL (XON/XOFF protocol); LSTETX.REL (ETX/ACK  protocol);
and LSTCEN.REL (a Centronics parallel printer driver used mainly in  Televideo
implementations).  Less  common  are LSTPAR.REL, the  driver  for  a  standard
Centronics  parallel printer; and LST300.REL, a simple, slow-speed,  Teletype-
like  serial driver. This has no provision for error detection, and relies  on
the  printer being able to keep up with the transmission rate. One or more  of
these drivers is placed in the GEN file of the appropriate server or satellite
processor  when  the GEN (system generation) program is run  to  generate  the
TurboDOS operating system.

2) Setting printer baud rates  (BR)
-----------------------------
After determining handshaking and selecting the appropriate driver module, the 
baud rate should be checked against the default value for that driver. If some
rate  other  than  the default is desired, use a text  editor  to  modify  the
appropriate  patch  point  in the PARameter file  containing  global  symbolic
patches  for the node. Examples of symbolic patch points are shown in  Listing
1.

Listing 1
---------
        Baud Rate patch point (PAR)  Printer driver (GEN)
        ---------------------------  --------------------
        CTSBR = hn ln   Default=6E           LSTCTS
        XONBR = hn ln   Default=07           LSTXON
        ETXBR = hn ln   Default=07           LSTETX
where  hn represents the high-order 'nibble' (4 bits), which can have  only  3 
valid values in these particular patch points:
     0  Represents the disabling of all hardware handshaking.
     4  Represents hardware handshaking.
        Bit 6 is set, and enables CTS.
        Is used with XONBR and ETXBR drivers to enable the respective protocol
        + CTS.
     6  Represents hardware handshaking for output only (input disabled).
        Is used for all CTSBR patch points.
and  ln  represents  the  low-order nibble, which can  have  16  valid  values 
representing the 16 baud rate values. The most common are:
        5 = 300
        7 = 1200
        E = 9600
Example: CTSBR = 6E     ;CTS handshaking at 9600 baud
         XONBR = 47     ;XON/XOFF + CTS handshaking at 1200 baud
         XONBR = 07     ;XON/XOFF handshaking at 1200 baud
         ETXBR = 05     ;ETX/ACK handshaking at 300 baud

Print spooling
--------------
The  2  print  modes  are  either  spooled  or  direct.  In  most   multi-user 
applications,  spooled  printing  is preferred over direct.  When  a  file  is
spooled,  TurboDOS  creates a print file on the disk. When the  print  job  is
done, TurboDOS will queue the print files for de-spooling in a first-in first-
out (FIFO) manner. Print file de-spooling is a background process that is done
automatically.
The  De-SPool Printer Assignment Table (DSPPAT), in the module LSTTBL,  is  an 
array  of 16 bytes (for printer A-P) that defines the queue assigned  to  each
printer.  Positions  1  through 16 in the array correspond  to  printers  A-P,
respectively.  The hex value (01-10, corresponding to printers A-P)  found  at
each  position  in  the  array defines the queue  to  which  that  printer  is
assigned.  A value of 00H indicates that the printer is off-line. The  default
value (01H) assigns all printers to queue A. A de-spool assignment table looks
like this:
DSPPAT = 01,01,02,02    ;Printer A to Queue A,
                        ;printer B to Queue A,
                        ;printer C to Queue B,
                        ;printer D to Queue B, etc.
The  files created by the spooler default to the system drive. To  change  the 
default, you may patch the symbol SPLDRV in the module LCLTBL, as follows:
SPLDRV = 0FFH   ;0FFH is default for system drive.
                ;Hex value of 0-F specify spool drive of A-P.
The print mode for a local user is determined by the symbol PRTMOD, located in 
module LCLTBL. The default value is 1, which specifies spooling. To change the
default, patch PRTMOD as follows:
PRTMOD = 1      ;1 is default for spooling.
                ;Hex value of 0 for direct printing,
                ;and 2 for print to console.

Print queues
------------
A  print  queue  is  a list of print  jobs  awaiting  de-spooling.  The  QUEue 
ASsignment  Table (QUEAST) defines which queues of A-P are local,  remote,  or
invalid. Also specified are the network addresses for each remote queue.
QUEAST = 00,(0000)      ;Queue A local   -- First 3 bytes zero
         0FF,(0000)     ;Queue B invalid -- First byte 0FFH
The  patch symbol QUEPTR in module LCLTBL specifies initial queue  or  printer 
assignments.  If  print  mode  is  spooled,  this  symbol  specifies  a  queue
assignment.  If  print  mode  is  direct,  this  symbol  specifies  a  printer
assignment.
QUEPTR = 1      ;1 is default.
                ;Hex values of 01-10 represent assignments of A-P.
                ;Zero signifies no queue, or printer off-line.

3) Printer assignment
---------------------
There  are 2 classes of printers in a TurboDOS network: local and remote.  The 
PrinTeR  ASsignment Table (PRTAST) contains, for each local or remote  printer
in  the  system, one byte specifying the printer, and 2 bytes  in  parentheses
specifying  the node address of the printer. In the printer designation  byte,
the  high-order nibble can only have 2 values: 0, signifying a  local  printer
physically attached to the node; or 8, signifying a remote printer attached to
some  other  node. The low-order nibble specifies the printer  number  (0-0FH,
corresponding to printers A-P). The following is an example of an entry:
        PTRAST = 00,LSTDRA,0FF,(0000),01,LSTDRA,85,(0002)
where:
        00,LSTDRA  -- Local printer A to channel 0
        0FF,(0000) -- Printer B is invalid
        01,LSTDRA  -- Local printer C to channel 1
        85,(0002)  -- Printer D is printer F on Remote node 2.
In  looking  at the first bytes (00, 0FF, and 01), we find  2  Local  printers 
attached  to this node, and one Remote (85). The high-order nibbles (0 and  8)
indicate  Local  and  Remote printers,  respectively.  The  low-order  nibbles
indicate  that  Printer  A is attached to channel 0 (first  serial  port)  and
Printer C is attached to channel 1 (second serial port).

Local printer configurations
----------------------------
Any  node  (whether  server or a satellite) can be configured  for  up  to  16 
printers, designated A-P. These printers can all be in use simultaneously, and
have  other  print jobs waiting in the queue. Before attempting to  work  with
full networking, we will look at local printer configurations.

Listing 2
---------
NODE.GEN  ;TurboDOS operating system GENeration file
LSTPAR          ;Printer driver for Parallel (Centronics) interface
LSTCTS          ;Printer driver for CTS hardware handshake
LSTETX          ;Printer driver for ETX/ACK software handshake
DSPOOL          ;Despooler for Local printer (only goes in satellites)

NODE.PAR  ;TurboDOS operating system PAR(ameter) file
CONAST = 00,CONDRA              ;First serial channel -- Console terminal
PTRAST = 00,LSTDRA              ;Parallel port -- No serial channel used
         01,LSTDRB              ;Second serial channel -- Second printer driver
         02,LSTDRC              ;Third serial channel  -- Third printer driver
DSPPAT = 00,01,02               ;No queue (Direct)
                                ;Second printer Queue A
                                ;Third printer Queue B
QUEAST = 00,(0000),00,(0000)    ;Queues A and B are Local queues
CTSBR = 6E                      ;CTS with handshake at 9600 baud
XONBR = 07                      ;XON without handshake at 1200 baud
        Figure  1. Local printer configuration with parallel CTS and  XON/XOFF 
        protocol
The example shown in Listing 2 has 3 printers on a node. The first printer  is 
Direct,  and  uses a parallel interface. The second printer is  Spooled,  with
hardware  handshaking  at  9600  baud. The third printer  is  Direct,  with  a
software  handshake  at 1200 baud. See Figure 1. Note that the  assignment  of
LSTDRA is a 2-byte value assigning a Local printer to the first printer driver
in  the  GEN file. The assignment of LSTDRB is for a Local printer  using  the
second printer driver in the GEN file, etc.
Note  that  the  parallel driver (LSTPAR) does not use a  serial  channel  for 
communication  with  the printer, and this must be explicitly  stated  in  the
PTRAST entry. Note, too, that the positions of the printer drivers in the  GEN
file  directly correlate to how printers are assigned in the table.  The  last
letter of LSTDR? tells the printer assignment table which driver to use in the
GEN -- i.e., LSTDR(A) uses the first printer driver in the GEN, LSTDR(B)  uses
the second printer driver, and LSTDR(C) uses the third printer driver.

Listing 3
---------
NODE.GEN  ;TurboDOS operating system GENeration file
LSTCTS          ;Printer driver for CTS hardware handshake
LSTETX          ;Printer driver for ETX/ACK software handshake
LSTXON          ;Printer driver for XON/XOFF software handshake
DSPOOL          ;Despooler for Local printer (only goes in satellites)

NODE.PAR  ;TurboDOS operating system PARameter file
CONAST = 00,CONDRA              ;First  channel -- Console terminal
PTRAST = 01,LSTDRB              ;Second channel -- Second printer driver
         02,LSTDRC              ;Third  channel -- Third printer driver
         03,LSTDRA              ;Fourth channel -- First printer driver
DSPPAT = 01,02,03               ;Printer A to Queue A, B to B, etc.
QUEAST = 00,(0000),00,(0000)    ;All Queues valid and Local
         00,(0000)
CTSBR = 6E                      ;CTS printer at 9600 baud
ETXBR = 0C                      ;ETX/ACK at 4800 baud
XONBR = 67                      ;XON/XOFF + CTS by setting high-order
                                ;nibble to 6, at 1200 baud.
        Figure 2. Local printer configuration with ETX/ACK, XON/XOFF plus  CTS 
        and high-speed CTS
The  example given in Listing 3 shows 3 printers on a single node.  The  first 
printer  is  Spooled, with software handshaking (ETX/ACK) at  4800  baud.  The
second  printer  is  Spooled,  with both a  software  and  hardware  handshake
(XON/XOFF and CTS) at 1200 baud. The third printer is Spooled, with a hardware
handshake at 9600 baud. See Figure 2.

Printer networking
------------------
TurboDOS, by its nature, is a networking operating system that networks via  a 
distributed   processing  architecture.  A  TurboDOS  circuit  is  a   network
communication  path between individual processor nodes. In a  single  computer
system, there is a simple and closely-coupled connection between the nodes. An
area  of  confusion to TurboDOS users is printer networking, partly due  to  a
lack  of detailed documentation. Listing 4 comprises some examples  that  will
illustrate networking applications.

Listing 4
---------
PTRAST = 00,LSTDRA,81,(0001)
where:
00  is  one byte consisting of a high-order nibble (Local=0, Remote=8)  and  a 
low-order  nibble  (Local=port or channel number, Remote=printer  A-P  in  Hex
values of 0-F).
LSTDRA  is a 2-byte assignment entry (global symbolic address of driver  entry 
point) for Local printers, in which the last substitution character points  to
the printer driver to use in the GEN file, e.g., A = First Driver, B = Second,
C = Third.
81 is a Remote printer, which is Printer B in Remote's PTRAST.
(0001) is the Remote assignment entry in which the first byte of 00 refers  to 
the  circuit,  and the second byte of 01 refers to the node  on  that  circuit
(Circuit 00 = server, Node 01 = First satellite).
For  a remote printer, the first byte must have the Sign bit set. To  set  the 
Sign  bit, the high-order nibble of the first byte must have the Hex value  of
8.  This  lets the Local Node know that this is a Remote printer, and  is  not
physically  attached.  The low-order nibble of the first  byte  specifies  the
printer  letter to be accessed on the Remote processor. The  'word'  following
the  first  byte  specifies  the network  address  of  the  Remote  processor,
consisting of a Circuit and a Node. When referring to hardware, a word for  an
8-bit processor is 8 bits and, for a 16-bit processor, is 16 bits.
When  referring  to words in TurboDOS (on both 8- and  16-bit  processors),  a 
single  word is 2 bytes or 16 bits. A word is specified in the PARameter  file
whenever  a  Hex  value  greater than 255 is entered, or  when  the  value  is
surrounded by parentheses.
Another printer assignment might be written this way:
PTRAST = 00,LSTDRA,0FF,(0000),01,LSTDRA,83,(0001),84,(0001)
This  assignment defines 2 printers physically attached to a Remote  satellite 
node.  Setting  the high-order nibble of the first byte to 8 tells  the  Local
node (server) that 2 other printers are Remote. The Local node (server) "sees"
them  as printers D and E, corresponding to the order of the printers  in  the
assignment  table. The low-order nibbles of the first bytes, with values of  3
and  4, respectively, tell the Local system to look to the printer  assignment
table  of the Remote node for D and E. The 2-byte entry (Circuit 00, Node  01)
tells  the Local processor that the Remote printers are attached to Node 1  of
the network (satellite 1).
Printer assignment of Remote node (satellite 1) would look like this:
PTRAST + 9 = 01,LSTDRA,02,LSTDRA
This assignment shows a 9-byte offset resulting from 3 printers being assigned 
to the system defaults on Node 0. Each printer on Node 0 (A, B, C) takes up  a
3-byte  entry;  thus, 3 bytes * 3 system printers = 9 byte  offset  in  Remote
PTRAST. Printer D is Local, and physically attached to serial channel 1, while
printer E is Local and physically attached to serial channel 2.
A  remote networking example is given in Listing 5; please, refer to Figure  3 
for a diagramatic representation. When configuring a network, it helps to draw
diagramatic  representations like the ones of Figures 1, 2, and 3, to  aid  in
visualizing the network.

Listing 5
---------
MASTER.GEN  ;TurboDOS operating system Master GENeration file
NETREQ          ;Network request module
MSGFMT          ;Message format tables for NETREQ
LSTCTS          ;Printer driver for CTS hardware handshake
LSTPAR          ;Printer driver for Parallel handshake

MASTER.PAR  ;TurboDOS operating system PARameter file
CONAST = 00,CONDRA              ;First  serial channel -- Console terminal
PTRAST = 01,LSTDRA              ;Second serial channel -- First printer driver
         00,LSTDRB              ;Parallel channel -- Second printer driver
         80,(0001)              ;Printer C is printer A on R-node#1
         80,(0002)              ;Printer D is printer A on R-node #2
QUEAST = 0,(0),0,(0)            ;Queues A and B are valid Local queues
         80,(0001),80,(0002)    ;Queues C and D are valid Remote queues
DSPPAT = 01,02,03,04            ;Printer A to Queue A, B to B, etc.
CTSBR = 6E                      ;CTS printer at 9600 baud

SLAVE1.GEN  ;TurboDOS operating system Slave 1 GENeration file
NETSVC          ;Network service local printer request
DSPOOL          ;Print despooler for local printer
LSTETX          ;Printer driver for ETX/ACK handshake

SLAVE1.PAR  ;TurboDOS operating system Slave 1 PARameter file
PTRAST = 01,LSTDRA                 ;2nd serial channel -- 1st printer driver
         0FF,(0),0FF,(0),0FF,(0)        ;Printers B, C, D are Invalid
QUEAST = 0,(0)                          ;Local Queue A is Valid
         0FF,(0),0FF,(0),0FF,(0)        ;Queues B, C, and D are Invalid
ETXBR = 0C                              ;ETX/ACK handshaking at 4800 baud

SLAVE2.GEN  ;TurboDOS operating system Slave 2 GENeration file
NETSVC          ;Network service local printer request
DSPOOL          ;Print despooler for local printer
LSTXON          ;Printer driver for XON handshake

SLAVE2.PAR  ;TurboDOS operating system Slave 2 PARameter file
PTRAST = 01,LSTDRA              ;Second serial channel -- First printer driver
         82,(0000)              ;Printer B is printer C on R-node #0
         0FF,(0),0FF,(0)        ;Printers C and D are Invalid
QUEAST = O,(0)                  ;Local Queue A is valid
         82,(0)                 ;Queue B is Queue C on R-node #0
         0FF,(0),0FF,(0)        ;Queues C and D are Invalid
DSPPAT = 01,02                  ;Printer A to Queue A,B to B, etc.

SLAVE3.GEN  ;TurboDOS operating system Slave 3 GENeration file
LST300          ;Serial driver default 300 baud
                ;NETSVC required because of no local printer

SLAVE3.PAR  ;TurboDOS operating system Slave 3 PARameter file (Example 1)
PTRAST = 80,(0000)              ;Printer A is printer A on R-node #0
         81,(0000)              ;Printer B is printer B on R-node #0
         01,LSTDRA              ;Printer C is a Local printer
         0FF,(0000)             ;Printer D is Invalid
QUEAST = 80,(0)                 ;Queue A is Queue A on R-node #0
         81,(0)                 ;Queue B is Queue B on R-node #0
         0FF,(0),OFF,(0)        ;Queues C and D are Invalid
DSPPAT = 01,02                  ;Printer A to Queue A, B to B
                                ;No baud rate specified, so it goes to default

SLAVE3.PAR  ;TurboDOS operating system Slave 3 PARameter file (Example 2)
PTRAST +6 = 01,LSTDRA,0FF,(0)   ;Offset of 6 bytes for first 2 printers A and B
                                ;Printers A and B default to the server PTRAST
QUEAST +6 = 0FF,(0)             ;6-byte offset for the 2 system printers off
                                ;the server.
            0FF,(0)             ;Queues C and D are Invalid
DSPPAT = 01,02                  ;Queue A to Queue A, B to B

SLAVE4.PAR  ;TurboDOS operating system Slave 4 PARameter file
This satellite does not require any printer drivers in the SLAVE4.GEN file. In 
the  SLAVE4.PAR  file, it will require no printer or  queue  assignment.  This
satellite  defaults  to the server, but a sample PAR file is  shown  below  to
illustrate the defaults.
PTRAST = 80,(0),81,(0)          ;Defaults for system printers
         82,(0),83,(0)
QUEAST = 80,0),81,(0)           ;Defaults Queues for system printers
         82,(0),83,(0)
        Figure 3. Remote printer networking

EOF
TDOS5.WS4
---------

- "The Networking Capabilities of TurboDOS"
  Michel Simon and William Poole (Commercial Dynamics)
  "Microsystems", August 1984, Vol.5, No.8, p.78

(Retyped by Emmanuel ROCHE.)


By  now,  most people in the CP/M world are, at least vaguely,  familiar  with
TurboDOS.  It is fairly well known that TurboDOS runs most off-the-shelf  CP/M
software,  and  that it can be found on S-100  Bus-based,  multi-user,  multi-
processor systems (networks). Less well known is the fact that a multitude  of
network configurations are possible with TurboDOS. Some manufacturers are  now
experimenting   with  different  network  architectures,  and  a   number   of
interesting products already exist. Given the availability of TurboDOS on  16-
bit processors and the growing list of manufacturers integrating TurboDOS with
their  hardware,  we should see the use of TurboDOS networks increase  in  the
days to come.

TurboDOS  is  truly a network operating system. This article  focuses  on  the
networking aspects of TurboDOS, including its capabilities and uses. First, we
will  describe  the  features of TurboDOS that distinguish  it  as  a  network
operating  system.  Second, we will discuss the  various  possible  networking
configurations.  Next,  we  will  address  the  limitations  of  TurboDOS   in
configuring  larger  networks.  The remainder of this  article  contains  some
general  guidelines for choosing a TurboDOS network, followed by a  review  of
TurboDOS networking products currently available. Detailed descriptions of the
network  architectures of 3 different manufacturers' products are included  as
examples.


What makes TurboDOS a network operating system?
-----------------------------------------------

The  single  most  important  characteristics of TurboDOS  --  one  which,  in
essence,  distinguishes  a network operating system  from  a  single-processor
operating  system  -- is that it permits transparent message  routing  between
physically-connected processors. This feature allows a request for a  *DEVICE*
(TurboDOS  devices are disk drives, printers, and queues) to be  automatically
routed,  in  logical form (see below), to the processor on  the  network  that
controls the device. Routing takes place between nodes on a circuit and, in  a
multi-circuit network, between circuits via common nodes.

TurboDOS  recognizes a 'world' that consists of 1 to 255  network  *CIRCUITS*,
each  of  which can have from 1 to 255 *NODES* connected to it. A  node  is  a
microcomputer (single-board, standalone, diskless, etc.), and a circuit is the
means  (hardware and software) by which nodes are connected. The S-100 Bus  is
currently the most widely-used transmission facility (hardware) for a TurboDOS
circuit,  although  there is a rapid growth in the use of Local  Area  Network
(LAN) circuits (RS-422/SDLC, ARCnet, Omninet, Ethernet, etc.) with TurboDOS. A
node can be connected to, and thus can transfer information between, more than
one circuit.

In  order  to illustrate and clarify what makes TurboDOS a  network  operating
system,  let us compare the way CP/M handles a resource request with  the  way
TurboDOS  handles  the same request. When the operator of a  single-user  CP/M
system makes a request for a resource, such as:

        A>DIR C:

the chain of events is as follows:

     1. CP/M's   command  line  interpreter  sees  a  request  for   directory
        information  from the C drive, and makes a *LOGICAL FUNCTION  REQUEST*
        (in CP/M, a BDOS system call) requesting this information.

     2. The BDOS makes the appropriate calls to the BIOS, which contains  disk
        drivers   and   has  access  to  information  about   the   particular
        characteristics of drive C.

     3. The  BDOS retrieves the information from the BIOS, and passes it  back
        to  the command line interpreter, which then displays the  information
        on the console.

All  along,  CP/M  assumes  that  the drive  is  attached  to  the  requesting
processor,  since  CP/M  does not recognize the existence  of  more  than  one
processor.

Under TurboDOS, however, the chain of events is quite different:

     1. The  TurboDOS  command line interpreter sees a request  for  directory
        information from the C drive, and makes a logical function request  to
        the TurboDOS file system, asking for this information.

     2. TurboDOS checks an internal table to see if drive C is attached to the
        requesting  processor  (i.e.,  is  a "Local" resource)  or  if  it  is
        attached to some other processor on the network (i.e., is a non-local,
        "Remote" resource).

     3. If  drive C is a local resource, the procedure followed is similar  to
        that followed by CP/M.

     4. If drive C is not a local resource, TurboDOS gets a *NETWORK  ADDRESS*
        from  its internal table. The address indicates the circuit  and  node
        numbers of the processor to which the resource is attached.

     5. TurboDOS  forms a *MESSAGE PACKET* consisting of the logical  function
        request,  the data, and the network address, and passes it on  to  the
        *CIRCUIT DRIVER* of the circuit indicated in the network address.

     6. The  circuit driver has access to all the information it  needs  about
        the  hardware characteristics of its physical  transmission  facility,
        and  transmits  the message packet to the processor indicated  in  the
        network address.

     7. At  the receiving processor, the same cycle is repeated: the  TurboDOS
        system  on  that processor checks to see if the resource is  local  or
        non-local (remote). If the resource is local, the receiving  processor
        carries out the request; if non-local (remote), the processor forwards
        the  message  packet. The requested information is returned,  via  the
        same  path,  to the originating processor. This processor  passes  the
        information to the command line interpreter, which displays it on  the
        console.

The  internal  tables mentioned in Events 2 and 4, which  are  called  *DEVICE
ASSIGNMENT TABLES*, are configured when the TurboDOS operating system on  each
processor  is  generated. Thus, the message-forwarding process  is  completely
transparent to the user, who enters a DIR command to get a directory listing.

In  the above scenario, it is important to note that messages are  transmitted
over  the network in the form of logical function requests, enabling  TurboDOS
to  meet  one of the fundamental requirements of a network  operating  system:
that  only  one processor recognizes the hardware characteristics of  a  given
device. For example, it is imperative that only one processor maintain a  disk
drive's  allocation  vector  or storage-used list;  otherwise,  assuring  data
integrity  in  a  multi-user environment would be impractical.  To  meet  this
requirement, TurboDOS messages are transmitted in a device-independent  format
until they arrive at their final destination.

Print spooling is accomplished through a network in the same manner as  drives
are  accessed: printer and queue requests are forwarded to the processor  that
controls the resource, just as drive requests are forwarded.


Other features
--------------

An  important  and related feature of TurboDOS is  its  modular  construction.
Networking hardware dependencies are isolated in circuit driver modules, while
peripheral hardware dependencies are isolated in device driver modules. Device
drivers  and  circuit  drivers  take  hardware-independent  instructions  from
TurboDOS,  and  execute them on the specific device or network  hardware  that
they control. In addition, modules of the operating system and device  drivers
are easily linked by the TurboDOS GEN command. Therefore, individual  versions
of the operating system can be configured to contain only those modules needed
to control the resources attached to a given processor on the network.

A  notable  consequence  of TurboDOS's modularity is that a  wide  variety  of
peripheral  devices  and  network hardware can  be  integrated  into  TurboDOS
systems  by  EOMs and system integrators. Device drivers can  be  replaced  or
updated as new disk and networking hardware becomes available.

Another feature of TurboDOS that deserves note is its support of network file-
and  record-locking. The processor that controls a given device keeps an  open
file  and  record list locally. When a logical function request for  a  locked
file  or  record is received, access to that file or record is denied,  and  a
return code indicating the error is sent back over the network.

TurboDOS's network file and record locking allows almost all single-user  CP/M
and   multi-user  MP/M  software  to  run  on  a  TurboDOS   network   without
modification. File lockout prevents single-user programs from corrupting files
when  they are simultaneously accessed by more than one user.  Record  lockout
allows simultaneous multi-user access to data files in an organized fashion.

The  above features combine to make TurboDOS a versatile operating system  for
configuring  a  wide  variety  of  networks  with  different  topologies   and
transmission  facilities.  We  will now classify and describe  some  of  these
network configurations.


Tightly-coupled networks
------------------------

As we mentioned earlier, the most common TurboDOS networks currently available
are  on  multi-user, multi-processor S-100 Bus-based systems,  including  IMS,
North  Star,  MuSYS,  and  Advanced Digital, to name but a  few;  for  a  more
complete list, refer to Table 1.

Table 1. TurboDOS networking products

Format: Manufacturers
        Tightly-coupled architecture
          Loosely-coupled architecture:
          Server-to-server
          Satellite-to-satellite
          Topology
          Maximum distance (feet)

(NOTE: In this table, K = kiloBAUDS, M = megaBAUDS.)

Advanced Computer Technology
S-100 Bus star
  (In development)

Advanced Digital
S-100 Bus star
  (In development)

Alspa
RS-422 800K bus, 2000 ft
  No
  No

Business Operating Systems
Parallel star, 10 ft
  RS-232 19.2K
  RS-232 19.2K
  Point-to-point
  300

California Computer Systems
S-100 Bus star
  No
  RS-232 9.6K
  Point-to-point
  300

Commercial Dynamics
S-100 Bus star
  No
  Differential 300K
  Bus
  1000

HM Systems, Ltd. (UK)
S-100 Bus star
  2.5M ARCnet
  No
  Ring
  2000

Intercontinental Micro Systems
S-100 Bus star
  2.5M ARCnet
  No
  Ring
  2000

Independent Business Systems
S-100 Bus star
  No
  RS-422 800K / RS-232 38.4K
  Point-to-point
  600 / 5000

Industrial Micro Systems
S-100 Bus star
  No
  Differential 307K
  Bus
  1000

JC Systems
2.4M Parallel bus, 600 ft
  No
  Parallel 2.4M
  Bus
  600

Litton Industries (Sweda Int'l.)
2M Token-passing LAN
  2M Token-passing LAN
  2M Token-passing LAN
  Ring
  2000

MuSYS
S-100 Bus star
  10M Ethernet
  RS-422 500K / RS-232 9.6K
  Ethernet bus / Point-to-point
  9000 / 900 / 300

NCR Corp.
No
  2M Omninet
  No
  Bus
  2000

North Star
S-100 Bus star
  No
  No

Philips (Netherlands)
Euro-Bus star
  2M Token-passing LAN
  No
  Ring
  2000

QDP Computer Systems
S-100 Bus star
  No
  No

Sierra Data Sciences
S-100 Bus star
  No
  No

Teletek
S-100 Bus star
  No
  No

Televideo
RS-422 800K star, 300 ft
  RS-422 800K
  No
  Point-to-point
  300


Networks  of  this type are referred to as *TIGHTLY-COUPLED*  networks.  In  a
tightly-coupled networks, disk resources are largely centralized, and all  the
processors are booted from one disk.

The  master processor is called a "server" because it services  requests  from
the satellites for access to the (many or all) global resources attached to it
(disks, printers, etc.). Slave processors are called "satellites" because this
is  a  more  accurate description of their function as part  of  the  network.
Although  their  primary function is to execute user programs,  and  they  are
booted  by the server and dependent on it for most operations, satellites  are
independent  processors  and,  in some cases, possess local  disk  or  printer
resources.

In  a  discussion of network configurations, it is  important  to  distinguish
between the hardware architecture of the transmission facility and the logical
topology.  For  example, in a typical S-100 Bus implementation of  a  TurboDOS
network,  the  S-100 Bus is the transmission facility, but the network  has  a
star  topology (see Figure 1). That is to say: even though all the  processors
are  physically  connected on the S-100 Bus, the  satellites  cannot  directly
communicate  with each other; all communication on the tightly-coupled,  S-100
Bus-based  network  is  controlled by  the  server.  Although  tightly-coupled
networks   are   predominantly   S-100   Bus-based,   other    tightly-coupled
architectures  have been implemented. Examples include Alspa, JC Systems,  and
Televideo (see Table 1).

        Figure 1. Tightly-coupled network, star topology


Loosely-coupled networks
------------------------

Networks  having  processors  that are independent of  each  other  for  basic
operations,  and disk resources that are distributed throughout  the  network,
are  referred to as *LOOSELY-COUPLED* networks. Loosely-coupled  networks  can
connect standalone, single-user systems or tightly-coupled networks.

Loosely-coupled networks of tightly-coupled systems require that at least  one
processor on each tightly-coupled network belong to both its internal  circuit
and  the  loosely-coupled  circuit (see Figure 2). Either the  server  of  the
tightly-coupled  circuit  or a satellite can be  the  dual-purpose  processor,
providing,  of  course,  that it has a physical means  to  transmit  messages.
Server-to-server  networks  typically require additional  hardware;  they  are
generally  considered desirable when the loosely-coupled network will be  used
for  high-volume  disk  access  and/or  chassis  expansion  (e.g.,  MuSYS  and
Intercontinental Micro Systems). Satellite-to-satellite networks use  hardware
already  on  satellite boards, and run at low-to-medium speeds  for  file  and
peripheral sharing (e.g., Commercial Dynamics, MuSYS and IBS).

Eight-bit single-user personal computers have not been widely networked  using
TurboDOS  because of memory considerations. Until recently, TurboDOS  did  not
support  more than 64K of memory, and a fully-configured  TurboDOS  (including
file  system and disk drivers) leaves only a 43K-to-45K TPA on a  64K  system.
With  TurboDOS now available on the Intel 8088/8086 family of  processors,  we
can  expect  to  see TurboDOS become a popular  environment  for  implementing
networks of 16-bit single-user machines, such as the IBM PC.

        Figure 2. Loosely-coupled networks


Gateways
--------

As  TurboDOS  networks  become available for a  wide  variety  of  processors,
*GATEWAYS*  will  become  increasingly important. Since  a  processor  can  be
connected to as many TurboDOS network circuits as it has circuit drivers  (and
corresponding transmission facilities), a gateway processor that has different
types  of  network  interfaces  can  interconnect  networks  using  dissimilar
hardware (see Figure 3). (ROCHE> "Internet" means "interconnected networks"...
This  article, describing it using 8-bit microcomputers, was written 12  years
before  Bill  Gates reluctantly provided "Internet Explorer"  under  Microsoft
Windows...)

Because  of the TurboDOS network's forwarding mechanism, each network  circuit
connected  by  a  gateway processor can be accessed by  nodes  throughout  the
entire network. If a processor is not itself connected to a given circuit, its
internal *NETWORK FORWARDING TABLE* tells it how to access the other circuits.
Network  forwarding  is part of the general  message-forwarding  mechanism  of
TurboDOS described earlier, and is transparent to the user.

As  TurboDOS  networks for IBM PC and PC-compatibles begin to appear,  we  can
expect  to  see  gateways implemented to connect  them  to  existing  TurboDOS
networks.  The introduction of banked memory support in Z80 TurboDOS  1.3  has
made  it both feasible and desirable to network 8-bit, as well as 16-bit,  PCs
together.

        Figure 3. Multiple networks linked by a gateway processor


TurboDOS limitations
--------------------

Any  TurboDOS  network  that  has fewer than 16 devices of  any  type  can  be
configured  to emulate a large computer system; no matter how the hardware  is
physically distributed, users have access to all network devices at all times.
However,  since  TurboDOS uses the CP/M device-naming  convention  (letters  A
through P), when there are more than 16 devices on a network, users may access
only  a  subset  of the network at any one time. While  16  devices  may  seem
sufficient,  it  is  easy to imagine a network that has  more  than  16  drive
volumes  (or  printers); it is also reasonable for network users  to  want  to
access most, if not all, of them.

This  situation can be handled in 3 different ways. First, the network can  be
restricted  to a maximum of 16 of each device. This limitation  is  reasonable
for small point-to-point systems (e.g., 2 systems communicating directly  over
a  dedicated wire) but is prohibitive for larger networks. Second,  one  might
change the device assignments of one or more processors. This has the  serious
disadvantage that TurboDOS has no mechanism for changing device assignments in
a  running  system;  device assignments can be made only when  the  system  is
generated. Thus, this scheme would entail generation of a separate version  of
the system for each set of device assignments; to change the assignments,  the
user would have to shut down the current system and reboot it with a different
version.  Since few end-users generate their own systems, this is not  a  very
practical  solution. Last, TurboDOS networking products can be  provided  with
utilities  that  allow device assignments to be edited and patched  while  the
system  is running, without reSYSGENing the operating system or  rebooting  it
(see the Commercial Dynamics example, below).


Choosing a TurboDOS network
---------------------------

If  you  already  owns a TurboDOS tightly-coupled  network,  your  choice  for
further  networking  products  will probably be limited  to  whatever  network
hardware  your manufacturer or dealer supports. However, if you have  not  yet
purchased  a system or are in a position to choose between different types  of
networks,  an  analysis of your applications is important to  determine  which
type  of  tightly-coupled and/or loosely-coupled network will best  suit  your
needs.

All  of  the  currently-available  tightly-coupled  networks  provide  similar
network performance. Tightly-coupled networks that use an RS-422  transmission
facility have a somewhat slower network transfer rate and, therefore, will not
perform  as  well  as  S-100 and Parallel Bus  systems,  especially  in  disk-
intensive  operations. The costs of most tightly-coupled systems are  similar,
so  your  choice will depend primarily on the other features of  the  TurboDOS
system, including disk storage, application software, dealer support, etc.

Loosely-coupled networks, on the other hand, differ significantly in price and
performance.  A  low-speed  network  can  cost less  than  $100  per  node  to
implement, while the highest speed system currently available costs more  than
$2000   per  node.  Loosely-coupled  network  transmission  rates  also   vary
significantly, anywhere from 9.6 Kbaud to 10 Mbaud per second.

Low-speed networks (9.6 Kbaud to 40 Kbaud) are primarily useful for occasional
file  transfers, low-volume printer sharing, and the transmission of  user-to-
user  mail and messages. They are generally unacceptable in  situations  where
large  volumes  of data must be transferred or more than a few nodes  use  the
network at the same time.

Medium-speed networks (100 Kbaud to 500 Kbaud) provide adequate throughput for
all  but the most intensive networking speeds. A network in this  speed  range
should  be able to load executable files, transfer large amounts of data,  and
have multi-user access over the network. Medium-speed networks are most  often
used in situations where network nodes use primarily local disk resources  and
only  go  to  the network for access  to  globally-shared  resources.  Sharing
printers  and  file transfers between a loosely-coupled network  of  otherwise
independent, tightly-coupled systems is an excellent application for a medium-
speed network.

High-speed networks (500 Kbaud and higher) are needed for good performance  in
situations  where  network  nodes make their primary disk  requests  over  the
network. A common use of high-speed networks is for chassis expansion; e.g., 2
separate tightly-coupled S-100 Bus networks sharing a single large hard  disk.
In  this  case, the system without the hard disk uses only a floppy  or  small
hard  disk to boot its processors and, from that point on, all  disk  requests
are transmitted over the high-speed network to the system with the large  hard
disk.

To determine whether a given network speed suits your specific needs, you must
first calculate, in bits, the average and maximum amounts of data that will be
transferred over the network simultaneously. Divide each of these figures into
the effective transfer rate of the network (which is usually 10 to 50  percent
of  the  manufacturer's  stated  transfer rate...)  to  determine  the  actual
transmission   time  under  average  and  maximum  conditions.   This   simple
calculation  should give you a reasonable estimate of how well a network  will
perform.


Trends
------

Given  the increasing availability of lower-cost networking hardware  and  the
decreasing  price  of hard disks, we expect to see large  multi-user  TurboDOS
networks configured somewhat differently than they are at present. S-100  Bus-
based  systems are usually loaded with as many users as they can hold  (10  to
20).  To  add more users, another S-100 Bus system is networked to  the  first
(provided  there is a networking product available), and the second system  is
loaded to its fullest potential. The major advantage of this approach is  that
it  offers the lowest cost per user; the disadvantage is that  contention  for
shared  network resources slows down all users on the system. In  most  common
applications,  better overall performance can be obtained by distributing  the
users  throughout  several smaller (4- to 8-user) systems. These  systems  can
then  be  networked together to provide all users with access to  all  network
resources.

The  arrival  of  inexpensive hard disks and  networking  products  will  make
distributed networks more common. The distributed approach also provides  more
protection against system failure: if one small network goes down, all  others
can still continue to operate. In a large, centralized system, a system  crash
can disable all users.


Sample architectures
--------------------

The  MuSYS product line provides examples of all of the most  common  TurboDOS
network  configurations.  Their  tightly-coupled  network  includes  satellite
processors with an S-100 Bus star topology. They also offer an Ethernet  board
set  that  connects tightly-coupled S-100 Bus star networks  into  a  loosely-
coupled  high-speed master-to-master bus network. A user will see very  little
difference in response time between making a disk request over this high-speed
master-to-master  network  and making the same request on a  local  disk.  For
lower-volume  applications, MuSYS provides both a medium-speed (RS-422) and  a
low-speed (RS-232) point-to-point, satellite-to-satellite network that uses  a
dedicated  satellite  processor board as a controller. It would not  be  cost-
effective  to  connect  more than 2 or 3 tightly-coupled  systems  with  these
latter 2 networks, since each system must contain one dedicated satellite  for
each of the other systems to which it is connected.

The  Televideo  808/816 network architecture is significantly  different  from
other  TurboDOS  networks. At the center of Televideo's  tightly-coupled  star
network  is a dedicated server processor that controls hard and floppy  disks,
printers,  and either 8 or 16 RS-422 ports. Each RS-422 port may be  connected
point-to-point  to a workstation, which is a terminal with  processor,  memory
and, optionally, a local floppy disk. Although the workstations can have local
disk  storage,  they  are  not  capable of booting  off  the  local  disk  and
connecting  to  the network at the same time; thus, the system is  a  tightly-
coupled network, even though the processors are physically distributed.

The Televideo network also allows point-to-point, master-to-master  networking
through the same RS-422 ports. As with MuSYS, an RS-422 port must be dedicated
to  each system that is networked. But, since the ports are provided with  the
initial purchase of the system, 2 Televideo 816 systems can be networked for a
very small incremental cost.

Commercial  Dynamics  provides  a  workable  solution  to  the  network   size
limitations  imposed  by  TurboDOS. They introduce the concept  of  a  network
*ENVIRONMENT*,  which  is  a  particular view of  a  network.  An  environment
consists of a set of logical-to-physical device assignments that specify which
physical  devices (disk drives, printers, queues) are accessed when a  logical
device is requested. Commercial Dynamics supplies utility programs to  create,
store, modify and activate the environments desired by the users.

Each  user  on a network can define any number of environments  by  running  a
menu-driven,  environment-editing utility. Environment definitions are  stored
in  files,  to  be  activated when they are needed  by  running  the  ACTIVATE
command,  and  specifying the environment to be activated. Although  only  one
environment  can be active at a time, a new environment can be activated  with
one command. Thus, a given user's view of the network can easily be  redefined
and  activated  without affecting any of the other users on the  network,  and
without regenerating the TurboDOS operating system. These utility programs are
supplied  with Commercial Dynamics' satellite processor and TurboNET  network,
and can also be obtained for any TurboDOS network.

Commercial  Dynamics  also  produces a satellite  processor  with  an  onboard
network  interface.  As  many as 16 tightly-coupled S-100  Bus  networks  that
contain at least one Commercial Dynamics satellite can be connected at  medium
speed  over  a S-100 Bus network. The Commercial Dynamics satellite can  be  a
user processor and a network interface at the same time; thus, loosely-coupled
networks  can  be  configured for a very low  cost.  The  Commercial  Dynamics
satellite  can  be  integrated into any S-100 Bus  TurboDOS  network,  and  is
currently running on IMS systems.


Summary
-------

A  wide  variety  of networking options are  available  with  TurboDOS,  using
different network topologies and networking hardware. TurboDOS's (ROCHE>  ????
Missing Text ???)


EOF

TDOS5.WS4
---------

- "The Networking Capabilities of TurboDOS"
  Michel Simon and William Poole (Commercial Dynamics)
  "Microsystems", August 1984, Vol.5, No.8, p.78

(Retyped by Emmanuel ROCHE.)


By  now,  most people in the CP/M world are, at least vaguely,  familiar  with
TurboDOS.  It is fairly well known that TurboDOS runs most off-the-shelf  CP/M
software,  and  that it can be found on S-100  Bus-based,  multi-user,  multi-
processor systems (networks). Less well known is the fact that a multitude  of
network configurations are possible with TurboDOS. Some manufacturers are  now
experimenting   with  different  network  architectures,  and  a   number   of
interesting products already exist. Given the availability of TurboDOS on  16-
bit processors and the growing list of manufacturers integrating TurboDOS with
their  hardware,  we should see the use of TurboDOS networks increase  in  the
days to come.

TurboDOS  is  truly a network operating system. This article  focuses  on  the
networking aspects of TurboDOS, including its capabilities and uses. First, we
will  describe  the  features of TurboDOS that distinguish  it  as  a  network
operating  system.  Second, we will discuss the  various  possible  networking
configurations.  Next,  we  will  address  the  limitations  of  TurboDOS   in
configuring  larger  networks.  The remainder of this  article  contains  some
general  guidelines for choosing a TurboDOS network, followed by a  review  of
TurboDOS networking products currently available. Detailed descriptions of the
network  architectures of 3 different manufacturers' products are included  as
examples.


What makes TurboDOS a network operating system?
-----------------------------------------------

The  single  most  important  characteristics of TurboDOS  --  one  which,  in
essence,  distinguishes  a network operating system  from  a  single-processor
operating  system  -- is that it permits transparent message  routing  between
physically-connected processors. This feature allows a request for a  *DEVICE*
(TurboDOS  devices are disk drives, printers, and queues) to be  automatically
routed,  in  logical form (see below), to the processor on  the  network  that
controls the device. Routing takes place between nodes on a circuit and, in  a
multi-circuit network, between circuits via common nodes.

TurboDOS  recognizes a 'world' that consists of 1 to 255  network  *CIRCUITS*,
each  of  which can have from 1 to 255 *NODES* connected to it. A  node  is  a
microcomputer (single-board, standalone, diskless, etc.), and a circuit is the
means  (hardware and software) by which nodes are connected. The S-100 Bus  is
currently the most widely-used transmission facility (hardware) for a TurboDOS
circuit,  although  there is a rapid growth in the use of Local  Area  Network
(LAN) circuits (RS-422/SDLC, ARCnet, Omninet, Ethernet, etc.) with TurboDOS. A
node can be connected to, and thus can transfer information between, more than
one circuit.

In  order  to illustrate and clarify what makes TurboDOS a  network  operating
system,  let us compare the way CP/M handles a resource request with  the  way
TurboDOS  handles  the same request. When the operator of a  single-user  CP/M
system makes a request for a resource, such as:

        A>DIR C:

the chain of events is as follows:

     1. CP/M's   command  line  interpreter  sees  a  request  for   directory
        information  from the C drive, and makes a *LOGICAL FUNCTION  REQUEST*
        (in CP/M, a BDOS system call) requesting this information.

     2. The BDOS makes the appropriate calls to the BIOS, which contains  disk
        drivers   and   has  access  to  information  about   the   particular
        characteristics of drive C.

     3. The  BDOS retrieves the information from the BIOS, and passes it  back
        to  the command line interpreter, which then displays the  information
        on the console.

All  along,  CP/M  assumes  that  the drive  is  attached  to  the  requesting
processor,  since  CP/M  does not recognize the existence  of  more  than  one
processor.

Under TurboDOS, however, the chain of events is quite different:

     1. The  TurboDOS  command line interpreter sees a request  for  directory
        information from the C drive, and makes a logical function request  to
        the TurboDOS file system, asking for this information.

     2. TurboDOS checks an internal table to see if drive C is attached to the
        requesting  processor  (i.e.,  is  a "Local" resource)  or  if  it  is
        attached to some other processor on the network (i.e., is a non-local,
        "Remote" resource).

     3. If  drive C is a local resource, the procedure followed is similar  to
        that followed by CP/M.

     4. If drive C is not a local resource, TurboDOS gets a *NETWORK  ADDRESS*
        from  its internal table. The address indicates the circuit  and  node
        numbers of the processor to which the resource is attached.

     5. TurboDOS  forms a *MESSAGE PACKET* consisting of the logical  function
        request,  the data, and the network address, and passes it on  to  the
        *CIRCUIT DRIVER* of the circuit indicated in the network address.

     6. The  circuit driver has access to all the information it  needs  about
        the  hardware characteristics of its physical  transmission  facility,
        and  transmits  the message packet to the processor indicated  in  the
        network address.

     7. At  the receiving processor, the same cycle is repeated: the  TurboDOS
        system  on  that processor checks to see if the resource is  local  or
        non-local (remote). If the resource is local, the receiving  processor
        carries out the request; if non-local (remote), the processor forwards
        the  message  packet. The requested information is returned,  via  the
        same  path,  to the originating processor. This processor  passes  the
        information to the command line interpreter, which displays it on  the
        console.

The  internal  tables mentioned in Events 2 and 4, which  are  called  *DEVICE
ASSIGNMENT TABLES*, are configured when the TurboDOS operating system on  each
processor  is  generated. Thus, the message-forwarding process  is  completely
transparent to the user, who enters a DIR command to get a directory listing.

In  the above scenario, it is important to note that messages are  transmitted
over  the network in the form of logical function requests, enabling  TurboDOS
to  meet  one of the fundamental requirements of a network  operating  system:
that  only  one processor recognizes the hardware characteristics of  a  given
device. For example, it is imperative that only one processor maintain a  disk
drive's  allocation  vector  or storage-used list;  otherwise,  assuring  data
integrity  in  a  multi-user environment would be impractical.  To  meet  this
requirement, TurboDOS messages are transmitted in a device-independent  format
until they arrive at their final destination.

Print spooling is accomplished through a network in the same manner as  drives
are  accessed: printer and queue requests are forwarded to the processor  that
controls the resource, just as drive requests are forwarded.


Other features
--------------

An  important  and related feature of TurboDOS is  its  modular  construction.
Networking hardware dependencies are isolated in circuit driver modules, while
peripheral hardware dependencies are isolated in device driver modules. Device
drivers  and  circuit  drivers  take  hardware-independent  instructions  from
TurboDOS,  and  execute them on the specific device or network  hardware  that
they control. In addition, modules of the operating system and device  drivers
are easily linked by the TurboDOS GEN command. Therefore, individual  versions
of the operating system can be configured to contain only those modules needed
to control the resources attached to a given processor on the network.

A  notable  consequence  of TurboDOS's modularity is that a  wide  variety  of
peripheral  devices  and  network hardware can  be  integrated  into  TurboDOS
systems  by  EOMs and system integrators. Device drivers can  be  replaced  or
updated as new disk and networking hardware becomes available.

Another feature of TurboDOS that deserves note is its support of network file-
and  record-locking. The processor that controls a given device keeps an  open
file  and  record list locally. When a logical function request for  a  locked
file  or  record is received, access to that file or record is denied,  and  a
return code indicating the error is sent back over the network.

TurboDOS's network file and record locking allows almost all single-user  CP/M
and   multi-user  MP/M  software  to  run  on  a  TurboDOS   network   without
modification. File lockout prevents single-user programs from corrupting files
when  they are simultaneously accessed by more than one user.  Record  lockout
allows simultaneous multi-user access to data files in an organized fashion.

The  above features combine to make TurboDOS a versatile operating system  for
configuring  a  wide  variety  of  networks  with  different  topologies   and
transmission  facilities.  We  will now classify and describe  some  of  these
network configurations.


Tightly-coupled networks
------------------------

As we mentioned earlier, the most common TurboDOS networks currently available
are  on  multi-user, multi-processor S-100 Bus-based systems,  including  IMS,
North  Star,  MuSYS,  and  Advanced Digital, to name but a  few;  for  a  more
complete list, refer to Table 1.

Table 1. TurboDOS networking products

Format: Manufacturers
        Tightly-coupled architecture
          Loosely-coupled architecture:
          Server-to-server
          Satellite-to-satellite
          Topology
          Maximum distance (feet)

(NOTE: In this table, K = kiloBAUDS, M = megaBAUDS.)

Advanced Computer Technology
S-100 Bus star
  (In development)

Advanced Digital
S-100 Bus star
  (In development)

Alspa
RS-422 800K bus, 2000 ft
  No
  No

Business Operating Systems
Parallel star, 10 ft
  RS-232 19.2K
  RS-232 19.2K
  Point-to-point
  300

California Computer Systems
S-100 Bus star
  No
  RS-232 9.6K
  Point-to-point
  300

Commercial Dynamics
S-100 Bus star
  No
  Differential 300K
  Bus
  1000

HM Systems, Ltd. (UK)
S-100 Bus star
  2.5M ARCnet
  No
  Ring
  2000

Intercontinental Micro Systems
S-100 Bus star
  2.5M ARCnet
  No
  Ring
  2000

Independent Business Systems
S-100 Bus star
  No
  RS-422 800K / RS-232 38.4K
  Point-to-point
  600 / 5000

Industrial Micro Systems
S-100 Bus star
  No
  Differential 307K
  Bus
  1000

JC Systems
2.4M Parallel bus, 600 ft
  No
  Parallel 2.4M
  Bus
  600

Litton Industries (Sweda Int'l.)
2M Token-passing LAN
  2M Token-passing LAN
  2M Token-passing LAN
  Ring
  2000

MuSYS
S-100 Bus star
  10M Ethernet
  RS-422 500K / RS-232 9.6K
  Ethernet bus / Point-to-point
  9000 / 900 / 300

NCR Corp.
No
  2M Omninet
  No
  Bus
  2000

North Star
S-100 Bus star
  No
  No

Philips (Netherlands)
Euro-Bus star
  2M Token-passing LAN
  No
  Ring
  2000

QDP Computer Systems
S-100 Bus star
  No
  No

Sierra Data Sciences
S-100 Bus star
  No
  No

Teletek
S-100 Bus star
  No
  No

Televideo
RS-422 800K star, 300 ft
  RS-422 800K
  No
  Point-to-point
  300


Networks  of  this type are referred to as *TIGHTLY-COUPLED*  networks.  In  a
tightly-coupled networks, disk resources are largely centralized, and all  the
processors are booted from one disk.

The  master processor is called a "server" because it services  requests  from
the satellites for access to the (many or all) global resources attached to it
(disks, printers, etc.). Slave processors are called "satellites" because this
is  a  more  accurate description of their function as part  of  the  network.
Although  their  primary function is to execute user programs,  and  they  are
booted  by the server and dependent on it for most operations, satellites  are
independent  processors  and,  in some cases, possess local  disk  or  printer
resources.

In  a  discussion of network configurations, it is  important  to  distinguish
between the hardware architecture of the transmission facility and the logical
topology.  For  example, in a typical S-100 Bus implementation of  a  TurboDOS
network,  the  S-100 Bus is the transmission facility, but the network  has  a
star  topology (see Figure 1). That is to say: even though all the  processors
are  physically  connected on the S-100 Bus, the  satellites  cannot  directly
communicate  with each other; all communication on the tightly-coupled,  S-100
Bus-based  network  is  controlled by  the  server.  Although  tightly-coupled
networks   are   predominantly   S-100   Bus-based,   other    tightly-coupled
architectures  have been implemented. Examples include Alspa, JC Systems,  and
Televideo (see Table 1).

        Figure 1. Tightly-coupled network, star topology


Loosely-coupled networks
------------------------

Networks  having  processors  that are independent of  each  other  for  basic
operations,  and disk resources that are distributed throughout  the  network,
are  referred to as *LOOSELY-COUPLED* networks. Loosely-coupled  networks  can
connect standalone, single-user systems or tightly-coupled networks.

Loosely-coupled networks of tightly-coupled systems require that at least  one
processor on each tightly-coupled network belong to both its internal  circuit
and  the  loosely-coupled  circuit (see Figure 2). Either the  server  of  the
tightly-coupled  circuit  or a satellite can be  the  dual-purpose  processor,
providing,  of  course,  that it has a physical means  to  transmit  messages.
Server-to-server  networks  typically require additional  hardware;  they  are
generally  considered desirable when the loosely-coupled network will be  used
for  high-volume  disk  access  and/or  chassis  expansion  (e.g.,  MuSYS  and
Intercontinental Micro Systems). Satellite-to-satellite networks use  hardware
already  on  satellite boards, and run at low-to-medium speeds  for  file  and
peripheral sharing (e.g., Commercial Dynamics, MuSYS and IBS).

Eight-bit single-user personal computers have not been widely networked  using
TurboDOS  because of memory considerations. Until recently, TurboDOS  did  not
support  more than 64K of memory, and a fully-configured  TurboDOS  (including
file  system and disk drivers) leaves only a 43K-to-45K TPA on a  64K  system.
With  TurboDOS now available on the Intel 8088/8086 family of  processors,  we
can  expect  to  see TurboDOS become a popular  environment  for  implementing
networks of 16-bit single-user machines, such as the IBM PC.

        Figure 2. Loosely-coupled networks


Gateways
--------

As  TurboDOS  networks  become available for a  wide  variety  of  processors,
*GATEWAYS*  will  become  increasingly important. Since  a  processor  can  be
connected to as many TurboDOS network circuits as it has circuit drivers  (and
corresponding transmission facilities), a gateway processor that has different
types  of  network  interfaces  can  interconnect  networks  using  dissimilar
hardware (see Figure 3). (ROCHE> "Internet" means "interconnected networks"...
This  article, describing it using 8-bit microcomputers, was written 12  years
before  Bill  Gates reluctantly provided "Internet Explorer"  under  Microsoft
Windows...)

Because  of the TurboDOS network's forwarding mechanism, each network  circuit
connected  by  a  gateway processor can be accessed by  nodes  throughout  the
entire network. If a processor is not itself connected to a given circuit, its
internal *NETWORK FORWARDING TABLE* tells it how to access the other circuits.
Network  forwarding  is part of the general  message-forwarding  mechanism  of
TurboDOS described earlier, and is transparent to the user.

As  TurboDOS  networks for IBM PC and PC-compatibles begin to appear,  we  can
expect  to  see  gateways implemented to connect  them  to  existing  TurboDOS
networks.  The introduction of banked memory support in Z80 TurboDOS  1.3  has
made  it both feasible and desirable to network 8-bit, as well as 16-bit,  PCs
together.

        Figure 3. Multiple networks linked by a gateway processor


TurboDOS limitations
--------------------

Any  TurboDOS  network  that  has fewer than 16 devices of  any  type  can  be
configured  to emulate a large computer system; no matter how the hardware  is
physically distributed, users have access to all network devices at all times.
However,  since  TurboDOS uses the CP/M device-naming  convention  (letters  A
through P), when there are more than 16 devices on a network, users may access
only  a  subset  of the network at any one time. While  16  devices  may  seem
sufficient,  it  is  easy to imagine a network that has  more  than  16  drive
volumes  (or  printers); it is also reasonable for network users  to  want  to
access most, if not all, of them.

This  situation can be handled in 3 different ways. First, the network can  be
restricted  to a maximum of 16 of each device. This limitation  is  reasonable
for small point-to-point systems (e.g., 2 systems communicating directly  over
a  dedicated wire) but is prohibitive for larger networks. Second,  one  might
change the device assignments of one or more processors. This has the  serious
disadvantage that TurboDOS has no mechanism for changing device assignments in
a  running  system;  device assignments can be made only when  the  system  is
generated. Thus, this scheme would entail generation of a separate version  of
the system for each set of device assignments; to change the assignments,  the
user would have to shut down the current system and reboot it with a different
version.  Since few end-users generate their own systems, this is not  a  very
practical  solution. Last, TurboDOS networking products can be  provided  with
utilities  that  allow device assignments to be edited and patched  while  the
system  is running, without reSYSGENing the operating system or  rebooting  it
(see the Commercial Dynamics example, below).


Choosing a TurboDOS network
---------------------------

If  you  already  owns a TurboDOS tightly-coupled  network,  your  choice  for
further  networking  products  will probably be limited  to  whatever  network
hardware  your manufacturer or dealer supports. However, if you have  not  yet
purchased  a system or are in a position to choose between different types  of
networks,  an  analysis of your applications is important to  determine  which
type  of  tightly-coupled and/or loosely-coupled network will best  suit  your
needs.

All  of  the  currently-available  tightly-coupled  networks  provide  similar
network performance. Tightly-coupled networks that use an RS-422  transmission
facility have a somewhat slower network transfer rate and, therefore, will not
perform  as  well  as  S-100 and Parallel Bus  systems,  especially  in  disk-
intensive  operations. The costs of most tightly-coupled systems are  similar,
so  your  choice will depend primarily on the other features of  the  TurboDOS
system, including disk storage, application software, dealer support, etc.

Loosely-coupled networks, on the other hand, differ significantly in price and
performance.  A  low-speed  network  can  cost less  than  $100  per  node  to
implement, while the highest speed system currently available costs more  than
$2000   per  node.  Loosely-coupled  network  transmission  rates  also   vary
significantly, anywhere from 9.6 Kbaud to 10 Mbaud per second.

Low-speed networks (9.6 Kbaud to 40 Kbaud) are primarily useful for occasional
file  transfers, low-volume printer sharing, and the transmission of  user-to-
user  mail and messages. They are generally unacceptable in  situations  where
large  volumes  of data must be transferred or more than a few nodes  use  the
network at the same time.

Medium-speed networks (100 Kbaud to 500 Kbaud) provide adequate throughput for
all  but the most intensive networking speeds. A network in this  speed  range
should  be able to load executable files, transfer large amounts of data,  and
have multi-user access over the network. Medium-speed networks are most  often
used in situations where network nodes use primarily local disk resources  and
only  go  to  the network for access  to  globally-shared  resources.  Sharing
printers  and  file transfers between a loosely-coupled network  of  otherwise
independent, tightly-coupled systems is an excellent application for a medium-
speed network.

High-speed networks (500 Kbaud and higher) are needed for good performance  in
situations  where  network  nodes make their primary disk  requests  over  the
network. A common use of high-speed networks is for chassis expansion; e.g., 2
separate tightly-coupled S-100 Bus networks sharing a single large hard  disk.
In  this  case, the system without the hard disk uses only a floppy  or  small
hard  disk to boot its processors and, from that point on, all  disk  requests
are transmitted over the high-speed network to the system with the large  hard
disk.

To determine whether a given network speed suits your specific needs, you must
first calculate, in bits, the average and maximum amounts of data that will be
transferred over the network simultaneously. Divide each of these figures into
the effective transfer rate of the network (which is usually 10 to 50  percent
of  the  manufacturer's  stated  transfer rate...)  to  determine  the  actual
transmission   time  under  average  and  maximum  conditions.   This   simple
calculation  should give you a reasonable estimate of how well a network  will
perform.


Trends
------

Given  the increasing availability of lower-cost networking hardware  and  the
decreasing  price  of hard disks, we expect to see large  multi-user  TurboDOS
networks configured somewhat differently than they are at present. S-100  Bus-
based  systems are usually loaded with as many users as they can hold  (10  to
20).  To  add more users, another S-100 Bus system is networked to  the  first
(provided  there is a networking product available), and the second system  is
loaded to its fullest potential. The major advantage of this approach is  that
it  offers the lowest cost per user; the disadvantage is that  contention  for
shared  network resources slows down all users on the system. In  most  common
applications,  better overall performance can be obtained by distributing  the
users  throughout  several smaller (4- to 8-user) systems. These  systems  can
then  be  networked together to provide all users with access to  all  network
resources.

The  arrival  of  inexpensive hard disks and  networking  products  will  make
distributed networks more common. The distributed approach also provides  more
protection against system failure: if one small network goes down, all  others
can still continue to operate. In a large, centralized system, a system  crash
can disable all users.


Sample architectures
--------------------

The  MuSYS product line provides examples of all of the most  common  TurboDOS
network  configurations.  Their  tightly-coupled  network  includes  satellite
processors with an S-100 Bus star topology. They also offer an Ethernet  board
set  that  connects tightly-coupled S-100 Bus star networks  into  a  loosely-
coupled  high-speed master-to-master bus network. A user will see very  little
difference in response time between making a disk request over this high-speed
master-to-master  network  and making the same request on a  local  disk.  For
lower-volume  applications, MuSYS provides both a medium-speed (RS-422) and  a
low-speed (RS-232) point-to-point, satellite-to-satellite network that uses  a
dedicated  satellite  processor board as a controller. It would not  be  cost-
effective  to  connect  more than 2 or 3 tightly-coupled  systems  with  these
latter 2 networks, since each system must contain one dedicated satellite  for
each of the other systems to which it is connected.

The  Televideo  808/816 network architecture is significantly  different  from
other  TurboDOS  networks. At the center of Televideo's  tightly-coupled  star
network  is a dedicated server processor that controls hard and floppy  disks,
printers,  and either 8 or 16 RS-422 ports. Each RS-422 port may be  connected
point-to-point  to a workstation, which is a terminal with  processor,  memory
and, optionally, a local floppy disk. Although the workstations can have local
disk  storage,  they  are  not  capable of booting  off  the  local  disk  and
connecting  to  the network at the same time; thus, the system is  a  tightly-
coupled network, even though the processors are physically distributed.

The Televideo network also allows point-to-point, master-to-master  networking
through the same RS-422 ports. As with MuSYS, an RS-422 port must be dedicated
to  each system that is networked. But, since the ports are provided with  the
initial purchase of the system, 2 Televideo 816 systems can be networked for a
very small incremental cost.

Commercial  Dynamics  provides  a  workable  solution  to  the  network   size
limitations  imposed  by  TurboDOS. They introduce the concept  of  a  network
*ENVIRONMENT*,  which  is  a  particular view of  a  network.  An  environment
consists of a set of logical-to-physical device assignments that specify which
physical  devices (disk drives, printers, queues) are accessed when a  logical
device is requested. Commercial Dynamics supplies utility programs to  create,
store, modify and activate the environments desired by the users.

Each  user  on a network can define any number of environments  by  running  a
menu-driven,  environment-editing utility. Environment definitions are  stored
in  files,  to  be  activated when they are needed  by  running  the  ACTIVATE
command,  and  specifying the environment to be activated. Although  only  one
environment  can be active at a time, a new environment can be activated  with
one command. Thus, a given user's view of the network can easily be  redefined
and  activated  without affecting any of the other users on the  network,  and
without regenerating the TurboDOS operating system. These utility programs are
supplied  with Commercial Dynamics' satellite processor and TurboNET  network,
and can also be obtained for any TurboDOS network.

Commercial  Dynamics  also  produces a satellite  processor  with  an  onboard
network  interface.  As  many as 16 tightly-coupled S-100  Bus  networks  that
contain at least one Commercial Dynamics satellite can be connected at  medium
speed  over  a S-100 Bus network. The Commercial Dynamics satellite can  be  a
user processor and a network interface at the same time; thus, loosely-coupled
networks  can  be  configured for a very low  cost.  The  Commercial  Dynamics
satellite  can  be  integrated into any S-100 Bus  TurboDOS  network,  and  is
currently running on IMS systems.


Summary
-------

A  wide  variety  of networking options are  available  with  TurboDOS,  using
different network topologies and networking hardware. TurboDOS's (ROCHE>  ????
Missing Text ???)


EOF