Phil Budne's guide to TOPS-10 evolution provides an overview of release dates and features for each version. Richard Cornwell's list of software kits shows what's available.

In thus series of blog posts we're going to focus on version 6.03 from 1977, running on the PDP-10 KA processor. We'll use Richard Cornwell's pdp10-ka simh emulator to run this.

TOPS-10 has a resident supervisor which has a dual role: firstly, providing access to machine resources and scheduling tasks, similar to a kernel on Unix based OSs. Its second role is to provide the main interface to user via the command control language - similar to a shell, but with code shared by users. Scheduling was done via priority queues, with support for real-time jobs.

The system can swap jobs in and out of disk and share code segments between programs. The particular version we are looking at did not support virtual memory and paging, however.

Users are identified by a project/programmer pair of numbers - eg [10, 11] - and access to the system is controlled by a password for each user.

The file system can span multiple disks. Each user gets their own directory to store files, and can share files by granting access to them. It's also possible for users to create sub-directories.

Peripheral devices such as printers, tapes or punched can either by shared by all users, with spoolers running to marshal access, or can be allocated to a specific user via a command.

The core system provides a set of utilities to manage files - both by individual commands such as RENAME and the powerful PIP (peripheral interchange program). Several editors were provided, from the simple SOS to the more complex TECO. Compilers and interpreters were available for BASIC, Fortran, COBOL and many other languages.

The system has a batch facility that allows jobs to be run at the same time as time-sharing users are logged in.

In this series of blog posts I plan to cover

• A quick start guide to getting TOPS-10 running under emulation.
• Files and directories
• Advanced use of TOPS-10 commands
• PIP
• Editors: SOS and TECO
• System users and creating new users
• Getting data into and out of the system.
• Batch commands
• The operator's console
• Installing TOPS-10 from scratch
• Programming under TOPS-10
• A look at each programming language supported.
• The DECUS software collections

Bitsavers has a wide collection of documentation: see the TOPS10 and TOPS10 Software Notebooks directories. Note that these cover different versions of the operating system, so not all features may work - check the date on each manual. A good place to start is 1974's Introduction to DECsystem-10 Software.

I welcome any questions or comments, and also especially any corrections if I have got something wrong. Please email me at [email protected] and I will add it here and update the main text.

The first computer in the family was the PDP-6 which used germanium transistors. It had a 36 bit word, up to 256kW of core memory and ran at approximately 250kHz. It was large (500-700kg), expensive ($300k) and was not very reliable. Only 23 systems were sold, but the architecture was well received by its initial customers such as MIT.

A very similar architecture but advances in digital logic technology and a reliable manufacturing process was used for the PDP-10. Several CPU models were introduced, starting with the KA10 in 1968. This was replaced by the KI10 using TTL and in 1975 by the ECL based KL10. A cheaper variant called the KS10 arrived in 1978. DEC marketed the PDP-10 under the name DECsystem-10 and -20. Xerox PARC and several startups also experimented with clones of the hardware.

During its lifetime, advanced in memory, storage and peripherals were introduced, and there were also multiprocessor variants.

DEC's original operating system for the PDP-6 was called the Monitor and used just 6kwords of memory. This was enhanced and renamed as TOPS-10 (Time-sharing / Total Operating System-10) with the introduction of the KA processor.

Sites would often customise TOPS-10 systems, adding special hardware and facilities to support them in the OS. One notable version was Stanford's WAITS, which evolved to be a separate operating system. The time-sharing bureau Tymshare developed TYMCOM-X based on TOPS-10.

BBN took the KA and added paging hardware which was not compatible with TOPS-10; they built their own system called TENEX. Eventually, DEC brought this in-house and released it as a supported OS called TOPS-20.

MIT, as an early PDP-6 customer for their AI lab, went a completely different direction, developing their own unique operating system called ITS (Incompatible Time-sharing System, a play of words on CTSS.

There's a great collection of PDP-10 software and documentation in the Github PDP-10 organisation, created by Lars Brinkhoff and around 50 other volunteers.

Software kits are available for TOPS-10, TOPS-20, WAITS and ITS: see the above, the trailing-edge collection of tapes, and Richard Cornwell's page.

There is a large amount of documentation on Bitsavers.

Zane Healy's The DEC PDP-10 Emulation Webpage also has a good collection of links to other PDP-10 sites, some of which you may need to use the Wayback Machine to access.

Probably the most comprehensive suite of emulators is Richard Cornwell's simh-based emulators for the PDP-6 and the KA/KI/KL/KS model PDP-10s. This had been merged into opensimh.

Also of note is Ken Harrenstein's KLH10 for the KS10 and KL10, and Bob Supnik's original KS10 emulator for simh.

The PiDP-10 recreates in hardware the PDP-10's front panel connected to a Raspberry Pi running simh. A lot of MIT AI Lab hardware is also demonstratable using this project.

The Interim Computer Museum holds a collection of PDP-10s formerly exhibited at Seattle's Living Computer Museum. Many of these systems, along with some emulators, are available online at ssh [email protected]:.

I plan to cover each of the below available operating systems, though probably not all immediately. When a system is added, click the links below to find out more.

• TOPS-10
• WAITS
• ITS
• TOPS-20

Wikipedia's page for the PDP-10 is a good place to start.

Gordon Bell, who was an architect at DEC for the PDP-6 and other systems, wrote a paper The Evolution of the DECsystem 10 which gives a great overview of the system's hardware and software, along with its impact.

I welcome any questions or comments, and also especially any corrections if I have got something wrong. Please email me at [email protected] and I will add it here and update the main text.

From the Algol source code of the language:

WIPL - WISCONSIN INTERACTIVE PROBLEM-SOLVING LANGUAGE
    BY  ED HARRIS  AND  BOB JANOSKI
    UNIV. OF WISCONSIN COMPUTING CENTER  (MADISON)
VERSION 1.50 (9/1/70)

So we know its name and source, but this is not much help - there is no documentation preserved and an internet search does not turn up much detail. I did find a UW report about a simulation program written in WIPL for school administration from 1969 at the Internet Archive. This includes an instruction manual for running the simulation, showing how to start WIPL and load the program - looks like this was run on the DCMCP rather than via CANDE. But no information on the language itself.

However, the CUBE tape also contains an online help file for the system which defines its syntax and how to use it. Looking at this, what stands out is its use of decimal line numbers and references to PART, for example in this fragment of a sample program:

3.0 TYPE "ENTER THE DEGREE OF THE POLYNOMIAL"
3.1 ACCEPT DEGREE
4.0 TYPE "ENTER THE COEFFICIENTS"
4.1 DO PART 30 FOR I=1 TO DEGREE+1

This makes me think this must be a JOSS clone.

JOSS - the JOHNNIAC Open Shop System - was an early time-shared language created at the RAND corporation in 1963. This ran on the one-of-a-kind computer JOHNNIAC and offered 20-30 users the ability to write simple programs that looked like this:

1.3 Do part 5 for i=1(1)100.
5.1 Type "Hello, Wikipedia!".
5.2 Type "This is JOSS.".

Many places implemented their own version of JOSS under different names (presumably because RAND held the copyright on the name JOSS) - for example BBN developed TELCOMP, DEC had FOCAL and the University of Pittsburgh had PIL. Each version made changes to the syntax but the core of the language was the same - interactive interpreter, decimal line numbers, simple loops and conditionals. The intended user was a scientist or engineer that wanted to do quick calculations but did not want to have to learn a full programming language.

First off, I assume you have the contents of the 'extras' tape installed: if not see this article.

You will need one additional file from the source CUBE tape. Get the file CUBEB13.BCD from Github and attach it to your emulator tape drive, either from the retro-b5500 tape window or via the simh command

at mt0 -r -f p7b /path/to/CUBEB13.bcd

Then switch to the operator's console, press Escape and type:

? LOAD FROM CUBEB13 ERRORS/CARDS; END

This may take a minute or so while the file is located on the tape.

Finally, you will need to rename some files. On the operator console, enter each of these commands:

? CHANGE WIPL/DISK TO 0WIPL/DISK; END
? CHANGE HELPFIL/CARDS TO WIPLS/HELPFIL; END
? ?CHANGE ERRORS/CARDS TO WIPLS/ERRORS; END

You can now run WIPL from CANDE via the RUN WIPL/DISK command. A quick example of starting WIPL, viewing the first help page and trying some immediate mode calculations before quitting and returning to CANDE:

RUN WIPL/DISK
 RUNNING
        

TYPE HELP IF YOU HAVE QUESTIONS  (WIPL VERSION 1.5)
?HELP
PLEASE WAIT.
      YOU ARE USING WIPL, A B5500 INTERACTIVE LANGUAGE DEVELOPED
      BY THE UNIVERSITY OF WISCONSIN COMPUTING CENTER.  TO OBTAIN
      INFORMATION ON THE USE OF ONE OF THE STATEMENT TYPES, TYPE
      "HELP" FOLLOWED BY THE STATEMENT TYPE, SUCH AS "HELP SET".
      IF YOU WOULD LIKE TO KNOW ABOUT ANY SPECIFIC AREA OF THE
      LANGUAGE, YOU MAY CHOOSE FROM THE FOLLOWING LIST OF TOPICS:
      (TO SELECT TOPIC 2.00 TYPE "HELP 2)

 2.00 GENERAL INFORMATION
 3.00 CALCULATOR MODE
 4.00 STORED PROGRAM MODE
 5.00 SPECIAL STATEMENTS
 6.00 DEBUGGING AIDS
 7.00 SAMPLE PROGRAM
?SET A = 2
?SET B = 40
?SET C = A + B
?TYPE C
 42 
?QUIT


 END WIPL .0 SEC.

You can also use SAVE and LOAD to store programs in the file system.

So with this, we can implement the TPK algorithm in WIPL.

1.000 REM TPK ALGORITHM IN WIPL
1.010 DIM A[11]
1.020 PRINT "PLEASE ENTER 11 NUMBERS"
1.030 DO PART 2 FOR J = 1 TO 11 BY 1
1.040 PRINT "RESULTS ARE"
1.050 DO PART 3 FOR J = 11 TO 1 BY -1
1.999 STOP

2.000 REM GET EACH NUMBER AND STORE IN ARRAY
2.010 ACCEPT A[J]

3.000 REM CALCULATE TPK AND DISPLAY RESULTS
3.010 RESULT = SQRT(ABS(A[J])) + 5 * A[J] ** 3
3.020 IF RESULT GTR 400 PRINT "TOO LARGE" ELSE PRINT RESULT

I think this is fairly easy to understand, but here are a few notes. WIPL (and JOSS) programs are separated into parts (the number before the decimal point), each of which may have several lines (the number after the decimal point). A part is like a subroutine with an implicit return when the next part starts. So here we have part 1 for the entry point. It calls part 2 using a loop to input numbers, and then part 3 with a reversed loop to calculate TPK.

Variables are floating point only. Arrays need to be dimensioned before use, and are 1-indexed.

Source and a transcript of its execution can be found on Github. I have also provided a version of the code in JOSS, so you can see how the syntax has changed for WIPL.

While researching this post I found this message by Paul Kimpel. Burroughs had their own version of JOSS, originally called BOSS and later renamed to INTERP. This is actually installed on disk as part of MCP Mark XIII set up. as INTERP/DISK. However, no documentation or help file is available. I can start it, but get the error message:

-INVALID USER INTERPS SAVFILE, S = 43, A = 37

suggesting there may be a step needed to grant access to the program. The source of INTERP is available on the MCP symbol tapes, however, so this would be an interesting project to understand the code and get this version of JOSS working.

Bitsavers has a collection of original RAND documents about JOSS. There is an ongoing project to restore RAND's PDP-6 version of JOSS-II at Github.

I welcome any questions or comments, and also especially any corrections if I have got something wrong. Please email me at [email protected] and I will add it here and update the main text.

• APL\B5500, an APL interpreter from the University of Washington adapted for time-sharing.
• WIPL, the (University of) Wisconsin Interactive Problem-solving Language - a JOSS clone.
• A LISP interpreter reported to come from Fort Monmouth, New Jersey.
• A 1970s era BASIC interpreter called XBASIC from the University of West Scotland

Read on for a brief look at APL. I will cover WIPL in a future article.

For LISP, I could get this running but it crashed on all input. XBASIC also runs but seems to require a tape to do anything useful. Let me know if you have success with these.

• EULER, an early language developed by Niklaus Wirth.
• PASCAL as implemented at Heriot-Watt University in around 1975.
• SNOBOL3 as implemented at the University of Wisconsin.
• GTL, the Georgia Tech Language, created at the university of the same name. An enhanced version of Algol 60.
• MCALGOL, an enhanced version of Burroughs Enhanced Algol, developed at Westinghouse.
• OMNITAB, a statistics language developed by NIST.

Apart from OMBITAB, which is in Fortran, these are all written in Algol. All were designed to run under batch (although CANDE could be customised to run them, as Georgia Tech did) so I will not look at these further.

As top level directories in the collection:

• R/C (Remote/Card), a remote job editor/submit package from Burroughs Defence.
• A Burroughs B6500 simulator and accompanying early MCP for the B5500
• Fragments of Mark-XIV and Mark-XVI MCP
• KRUNCH and UNKRNCH utilities - compress your Algol card deck by removing whitespace and comments.

The CUBE tape has many Algol source code routines. mostly for mathematical/engineering problems, a computer aided teaching system from GT, ELIZA, a tic-tac-toe game, an Algol to Fortran translator written in GTL and many other small programs. The library index is a good place to start exploring these.

Finally, the archive contains a few more recently developed programs

• A Sudoku solver by Paul Cumberworth
• Burroughs B5500-related utility programs for "Emode"
• Utilities from Paul Kimpel and Richard Cornwell

All of the above is in source form. To get them individually installed on your system you will need to compile them and install associated files as per the techniques described in the batch and data transfer posts.

However, a much easier way has been provided by Richard Cornwell, who has made a virtual tape file containing compiled files for many of the above. This is included in the simh distribution and can also be installed on retro-b5500 by getting extras.bin from here, mounting it on the tape drive as a BCD file and entering the command

? LOAD FROM EXTRAS =/=; END

on the console. The file extra.job at the same location also contains a script for building this tape if you want to find out how each component was constructed.

This implementation was created in Burroughs Extended Algol at the Computer Science Group at the University of Washington by Gary Kildall, Leroy Smith, Sally Swedine and Mary Zosel. Gary Kildall became famous later as the creator of the CP/M operating system. The source code was originally found as a printed listing and transcribed to a source file; this was supplanted later by a digital copy found in the CUBE tapes.

This APL was designed to be as compatible as possible with the original APL for the IBM System/360. As typewriters with the APL glyphs did not exist for the Burroughs, alphanumerical transliterations are used, eg Δ is written as $ and ⍴ as RHO.

If you install the contents of the extras tape as described above, you can then start APL from a CANDE prompt by typing RUN APL/DISK. Below is a sample session where I start APL, use calculator mode to do some work, then quit and return to CANDE.

RUN APL/DISK
 RUNNING
        
      APL/B5500 UW COMPUTER SCIENCE # 3-05-71 
      
 LOGGED IN THURSDAY 11-05-25  06:33 
      
X := 30
Y := 12
X + Y
42
      
)OFF
      END OF RUN

 END APL .0 SEC.

I plan to add a project-tpk run of TPK using APL at a later date.

Created by Joseph Weizenbaum at MIT in 1964-6, ELIZA could be called the world's first AI chatbot. I was especially interested to see a version for the Burroughs B5500 having worked on the reconstruction of the CTSS/MAD original.

This version is written in Georgia Tech Language which was mentioned above. GTL incorporated Lisp features into Algol, and the code looks like it was a port of Cosell's Lisp version rather than Weizenbaum's original. The source code is dated 1970-1 and was written by Charles Fricks and John Covert, presumably at Georgia Tech..

Assuming you have installed the extras tape as per the previous section, you will have three files on your system

• ELIZA/SYMBOL - source code
• ELIZA/SCRIPT - run time script file
• ELIZA/ELIZA - executable

In order to run under time-sharing, you will need to rename the executable as CANDE expects runnable code to have a 0 as its first character. Switch over to the operator console, press Escape and type:

? CHANGE ELIZA/ELIZA TO 0ELIZA/ELIZA; END

ELIZA can now be run under CANDE by typing RUN ELIZA/ELIZA. After it starts up, you can interact with ELIZA by typing single lines of text. Type *QUIT to exit. Here's a sample run, using the conversation in Weizenbaum's 1966 ACM paper:

RUN ELIZA/ELIZA
 RUNNING
        

ONE MOMENT PLEASE...
HELLO. HOW ARE YOU TODAY?
?MEN ARE ALL ALIKE.
HOW?
?THEY ARE ALWAYS BUGGING US ABOUT SOMETHING OR OTHER.
WHAT INCIDENT ARE YOU THINKING OF?
?WELL, MY BOYFRIEND MADE ME COME HERE.
DOES THAT SUGGEST ANYTHING ELSE OF YOURS?
?*QUIT
GOODBYE


 END ELIZA .0 SEC.

It has some interesting features that distinguish itself from the original ELIZA:

1. It saves the recall information to a file for use in subsequent sessions.
2. After a certain amount of CPU time is used it will automatically quit with:

I AM VERY SORRY, BUT YOU HAVE USED UP YOUR TIME.

I welcome any questions or comments, and also especially any corrections if I have got something wrong. Please email me at [email protected] and I will add it here and update the main text.