IBM PC XENIX

In 1984, IBM briefly flirted with XENIX, Microsoft’s variant of UNIX licensed from AT&T.  Around 1983-1984, Microsoft and Intel worked on porting XENIX to the 286 processor; Intel shipped XENIX with a number of its development systems in the mid-1980s.

A good description of IBM’s flavor of XENIX may be found in the IBM Personal Computer Seminar Proceedings, Volume 2, Number 9, published in November 1984. The OS/2 Museum recently obtained a copy of this booklet, which is now being made available in PDF format.

XENIX may have been the first multi-tasking, multi-user, protected-mode operating system available for the IBM PC/AT, and for the PC family in general. However, it was neither the first nor the last UNIX variant which IBM made available for the IBM PC. The first was PC/IX (developed for IBM by INTERACTIVE Systems Corporation, or ISC), designed to run on the PC/XT. XENIX was in turn followed by AIX (likewise developed by ISC), first available on the RT PC, and later on PS/2 and RS/6000 systems.

The IBM PC XENIX was based on AT&T’s UNIX System III (rather than the newer System V) with a number of BSD enhancements. The latter included vi, C shell, and termcap/curses. Microsoft also implemented several enhancements, such as the ability to read and write DOS file systems.

Much like other Xenix variants before and after, IBM PC XENIX was optionally shipped with Text Formatting System (TFS) and Software Development System (SDS) packages. The TFS was based on the classic nroff/troff formatters and associated utilities. The SDS included a compiler, assembler, debugger, and assorted development tools like SCCS and make.

One of the Microsoft-specific enhancements was the ability to cross-compile to DOS using the SDS. The XENIX to DOS cross-development capability was likely frequently used internally at Microsoft in the mid-1980s.

The IBM PC XENIX came with rudimentary networking capabilities, namely uucp and micnet, though little is known about their specifications.

The IBM PC XENIX was hardly a successful operating system, and is much less well known than the Xenix variants sold by Microsoft/SCO. It is likely that IBM had little incentive for a strong UNIX push, as AT&T was at the time considered a potentially very serious competitor (which it turned out not to be, but that’s a different story).

This entry was posted in IBM, Microsoft, Xenix. Bookmark the permalink.

25 Responses to IBM PC XENIX

  1. Andreas Kohl says:

    In April 1984 IBM announced the XENIX OS for it’s m68k-based System 9000. This one was based on MS XENIX 2.3 and UNIX V7, for more details see Announcement 284-159.

    The IBM PC XENIX OS Version 1.0 was announced in August 1984 (284-279) for IBM PC AT 5170 Model 99 and should had become available at Q1 1985. So it’s basically an OEM version of XENIX 286 implementing a subset of UNIX System III (called XENIX System III) which has been ported by the Santa Cruz Operation.
    IBM PC XENIX OS Version 2.0 announced in March 1986 (286-115) was based on a subset of UNIX System V.
    This offerings were later replaced by IBM’s AIX PS/2 (developed by Locus Computing Corp. link) and SCO’s XENIX, UNIX System V/386 or Open Desktop.
    I’m not sure about IBM PC/IX development done by Microport or Locus?

  2. michaln says:

    Do you have some reliable indication that SCO was significantly involved in the IBM PC XENIX development (i.e. Xenix 286) or is it just a supposition? As far as I know, that port was mostly done by Microsoft with some help from Intel, but I could be wrong. IBM only ever mentioned Microsoft, and there is no mention of SCO whatsoever on the actual IBM PC XENIX installation disks (only IBM and Microsoft).

    As for AIX, check the Wikipedia entry for AIX and ISC – AIX was developed by ISC. Locus probably worked on the 386 port of AIX and certainly contributed the DOS Merge facility (i.e. DOS boxes). Basically it looks like Locus did the 386 and S/370 ports of AIX, but did not develop AIX per se.

    Same for PC/IX. There’s a reason why ‘IX’ stood for ‘Interactive Executive’ and it wasn’t because it had been developed by Locus or Microport…

    ISC itself of course ended up being bought by Kodak and then by Sun, and a few bits of ISC UNIX ended up in the Intel Solaris port.

  3. Andreas Kohl says:

    So I have to look for an article from a Tandy programmer.
    PC Xenix 1.0 could be a Microsoft Xenix 3.0 especially for 286? Microsoft never offered a 286 or 386 version.

    MS Xenix versions through 2.4 were based on Unix V7 (Altos, Tandy TRS-XENIX).
    MS Xenix ver. 3 was based on System III.
    System V based Xenix 286 was a complete development under the control of SCO.
    Yet 1981-12-08 Microsoft and SCO signed a letter of intent for SCO to be a second source.
    SCO released their port (later called XENIX-86) to the IBM PC in September 1983.

  4. michaln says:

    IBM PC XENIX was indeed Microsoft XENIX 3.0 (judging from the strings in the files). There are numerous “Copyright 1984” messages which likely make it related to the Intel XENIX 286 releases from 1984. MS XENIX 3.0 being based on System III also exactly matches the IBM document saying the IBM PC XENIX is a System III derivative.

  5. Andreas Kohl says:

    InfoWorld Vol.5 Issue 52 has an article starting at page 40 (Can UNIX ever fit personal computers?).
    Intel Xenix was for Multibus systems not IBM PC compatible ones – a quite different architecture.

  6. Andreas Kohl says:

    SCO Xenix documentation mentions copyrights by Microsoft Corp. from 1980… and by Santa Cruz Operation from 1983 …
    The manuals are from SCO XENIX System V (286) 2.1 and 2.2 and SCO XENIX System V/386 2.3.
    Microsoft only did the original Xenix (1980) for DEC PDP and possibly the port to Zilog Z-8000.

  7. michaln says:

    Yes, a quite different architecture… except for the CPU. So where did Intel’s 286 XENIX come from in 1984? It doesn’t look like it was from SCO.

    None of the links you posted show that SCO had a 286 XENIX in 1984, but Intel/Microsoft did. Sure, SCO had 8086 XENIX at that point… a quite different architecture from the 286.

    The timeline in the Tandy newsgroup posts is very fuzzy, and indicates that SCO started a 286 port after the IBM PC/AT became available, but Microsoft and IBM had clearly been working on it before then, and Intel had a released product in 1984. See the MS XENIX 3.0 press release (from 1983 as far as I can tell) and the Intel XENIX 286 overview from November 1984.

  8. Andreas Kohl says:

    Doug Michels mentions IBM PC Xenix:
    http://www.youtube.com/watch?v=xZpaISQ82vA

  9. Andreas Kohl says:

    The file called Press Release is in fact a scanned page (5) from MS Systems Journal (possibly November 1984).

  10. michaln says:

    If Microsoft is to be believed, Volume 1, Issue 1 of the Microsoft Systems Journal was published in October 1986.

  11. Andreas Kohl says:

    So from which publication else? I remember MS Press was starting in 1984 publishing books. So maybe some catalogue or developer newsletter?

    Some excerpts from Microsoft’s own view of history:
    “August 25, 1980

    Microsoft announces XENIX OS, a portable, UNIX-based operating system for 16-bit microprocessors.”

    “December 8, 1981

    Microsoft announces that is has signed a letter of intent to enter a second-source agreement with The Santa Cruz Operation, Inc. for the XENIX operating system. The agreement will provide additional services to purchasers of XENIX, especially pre- and post-sale support, maintenance, and documentation.”

    Source:
    http://channel9.msdn.com/series/history/the-history-of-microsoft-1980

  12. Andreas Kohl says:

    Important companies with real developers in the early Xenix years were Altos, HCR (Kanada) and Logica Ltd (England). Altos became Acer’s server business unit – they still call their models “Acer Altos”. Also I remember an Extended Acer File System (EAFS) for SCO UNIX/ODT/OSR. The other two companies were taken over by SCO.

  13. michaln says:

    There’s no question that SCO completely took over XENIX development after 1987 or so, once Microsoft completely lost interest in UNIX (but MS at the same time owned a chunk of SCO).

    The question is just how much SCO did before then, and what Microsoft did themselves – you probably won’t try to tell me that Microsoft didn’t do any XENIX development in the 1984-85 timeframe…

    The only fact I have at the moment is that the IBM PC XENIX disks contain zero mentions of SCO, while all the SCO releases I’ve seen are chock full of SCO copyright messages. Unfortunately the oldest SCO version for the PC that I have is from 1986, so in theory it’s possible that something changed between 1984-5 and 1986, although I don’t know what or why.

  14. Wasn’t SCO spun out of Microsoft when they lost interest? And they were some kind of independent unit as they weren’t in Redmond or Albuquerque..?

    Just like there apparently was a Xenix for the PDP-11, although I’ve only seen a brief mention about it on usenet.. And the MS Systems Journal #2 from 1987 goes on about Xenix/386 …

    Does anyone know what they ran on the MS Vax’es that apparently was unix based for their inhouse email?

  15. michaln says:

    No, SCO wasn’t spun out of Microsoft… it was an independent company started in 1979 which initially worked with Microsoft, and eventually completely took over XENIX development (but Microsoft owned a 25% stake in SCO). They’d always been in Santa Cruz as far as I know.

    There was supposedly even XENIX running on Sun workstations (strictly MS internal), and Microsoft’s VAX systems my have run XENIX too. Not a lot of information out there, and even less information that’s reliable.

  16. Andreas Kohl says:

    It’s not easy to find references nowadays. On USENIX 1983 two MS engineers were presenting TRS-XENIX on Tandy – but nothing about Intel. Perhaps former SCO engineer Ron Record (http://ronrecord.com/jobs.html#Employment) should know more and better of technical details? So why not ask him?

    It’s strange that IBM PC Xenix only supports up to 3 MB memory and max. 3 users (2 connected by serial lines). Late SCO Xenix 286 supported 16 MB and more users. Some companies here in Germany used Pentium/Celeron boxes with 32 MB (only half was used) in late 1990ies to migrate from older Xenix servers with success.

    Yes, Microsoft’s mail server was running XENIX on DEC VAX (as I can see from old posts by Gordon Letwin).

  17. michaln says:

    Actually, the 3 users/3MB limitation is very easy to explain. 3MB was the initial maximum supported memory size on the PC/AT – not an architectural limitation, just what IBM offered at a time. 3 users were 1 console + 2 serial ports. Again just what IBM offered at the time.

    It is doubtful that XENIX actually imposed those limitations, it’s just that IBM wasn’t willing to claim higher values when they didn’t offer more memory and serial ports.

  18. Andreas Kohl says:

    I know about the limitations of IBM PC AT 5170 Model 99 (1984) and later Model 239 (announced 1985-10-01 / 185-116):
    “XENIX 1.0 customers operating with a 5170 model 239 will
    require an update to the XENIX 1.0 installation diskette. This
    diskette is available through your IBM Authorized Personal Computer
    point of purchase. The IBM Personal Computing Assistance Center
    (PCAC) will distribute the XENIX 1.0 update upon request from
    qualified customers.”

    Later 5170 Model 339 (announced together with model 319 on 1986-04-02) supports up to 15.5 MB. IBM wrote: “IBM Personal Computer XENIX (TM) Operating System users require XENIX Version 2.0.” But V2.0 only supports 3 MB.

    It seems PC XT 286 (5162-286) that supports up to 12.6 MB memory can run XENIX too – again: “IBM Personal Computer XENIX (TM) Operating System users require XENIX Version 2.0.”

  19. Antoni Sawicki says:

    Just happen to have it in my collection 😉
    http://www.tenox.tc/tmp/ww/msxenix/pics/ibm-xenix.jpg

  20. Rachel says:

    I have a copy of the os with book anyone have a clue what the value of it is

  21. Michal Necasek says:

    That’s a good question. Putting it up for sale on eBay would answer it 🙂 I’d expect something in the 50-100 dollar range depending on what exactly you have and in what condition. I don’t suppose you have photos? The OS/2 Museum might be interested.

Leave a Reply

Your email address will not be published. Required fields are marked *