• Please review our updated Terms and Rules here

History behind the disk images of AT&T UNIX System V Release 4 Version 2.1 for 386?

A

atdt916

New Member
Joined
Jan 7, 2019
Messages
4
History behind the disk images of AT&T UNIX System V Release 4 Version 2.1 for 386?

I'm curious as to the origins of the AT&T UNIX System V Release 4 Version 2.1 binaries and source code that are floating around the internet. From what I've read, all System V UNIX releases were sold by vendors and AT&T never released their own version. Is this incorrect? My understanding was that to obtain a UNIX OS in the 80s and early 90s, one would need to get a compiled version from a vendor, such as HP-UX, AIX, Solaris/SunOS, etc. These companies would license SVR4 and then release their own versions--usually on their own hardware.

Given this, I wanted to find an early version of UNIX that ran on x86 and had little modifications done to it by the vendor. This lead me to searching for Dell UNIX, which I've read had very little, if any, changes made by Dell and was used as the reference UNIX implementation by Intel. However, much to my surprise during this search, I found YouTube videos of AT&T UNIX System V Release 4 Version 2.1 OS running on a white box 386. From the video, there was no vendor branding at all nor any copyright messages of another vendor, so it would be appear to be as pure a SysV R4 system as one can get. The source and RAW floppy images are even available from abandonware archives.

Where did this version come from and what is the history behind it? When did it become available? Was it the version sent to vendors prior to them taking it for their customizations? An internal AT&T build? Was it intended for any particular PC or set of PCs? I've seen no mention of this release in any wikis nor in magazine articles, so I'm curious where this release came from and when it came out.
 
atdt916 said:
From what I've read, all System V UNIX releases were sold by vendors and AT&T never released their own version. Is this incorrect?
Click to expand...

It has to be. It's certainly not true for the 3B2 (the SVR3 & 4 base development platform). I don't believe it's true for the WGS/386 release considering this is a photo I just took:

svr3-386.jpg
 
eeguru said:
It has to be. It's certainly not true for the 3B2 (the SVR3 & 4 base development platform). I don't believe it's true for the WGS/386 release considering this is a photo I just took:
Click to expand...

Yes - we ran a bunch of AT&T WGS386's as point-of-sale systems, and we had AT&T branded UNIX media for them that looked just like this photo. I may even have some meda someplace...
 
I was under the impression that the standard AT&T release medium for SVR4 and preceding was 9-track magnetic tape. It certainly was the case with System III.
 
atdt916 said:
From what I've read, all System V UNIX releases were sold by vendors and AT&T never released their own version. Is this incorrect? My understanding was that to obtain a UNIX OS in the 80s and early 90s, one would need to get a compiled version from a vendor
Click to expand...

It gets very confusing for 80386 System V because there are AT&T SVrX releases, and System V/386 releases, which I think were all based on SVr3 and the port may have been done by Interactive Systems.
I saw some language in the V/386 r 3.1 documents I just scanned which said the device drivers were from IS.

I also thought the reference platform was VAX up until SVr4, where they switched to ATT 3B2.

I tried to make sense of it as I was trying to dig up the documentation for the various System V releases, and never really did.

There is probably a real history for it, somewhere..

There was a source release for 386 unix on cartridge tape. I got excited when a copy showed up on eBay. When I got it I discovered that the support floppies were
there, but the cartridge tape was missing from the slip case. I also just bought a copy of 386Vr3.1 and the first two disks of the base system and tools weren't there.
 
Al Kossow said:
There was a source release for 386 unix on cartridge tape. I got excited when a copy showed up on eBay. When I got it I discovered that the support floppies were there, but the cartridge tape was missing from the slip case. I also just bought a copy of 386Vr3.1 and the first two disks of the base system and tools weren't there.
Click to expand...

Hm. I have a QIC-only copy of "System V/386 Release 3.2u Version 2.1". I've been trying to find the support floppies to match for ages.
 
eeguru said:
It has to be. It's certainly not true for the 3B2 (the SVR3 & 4 base development platform). I don't believe it's true for the WGS/386 release considering this is a photo I just took:

View attachment 50506
Click to expand...

Ah, very cool, thank you! So it would appear then that these Release 4 Version 2.1 disk images likely originate from being included with one such system. This past weekend I had spent a lot of time diving into the history of UNIX and tracing the release lineage of the various versions. It's a shame that these releases are not as popular as I think they should be. Either that, or I've simply been looking in the wrong places these past 48 hours. In my opinion, having an AT&T SVR4v2.1 release included with an AT&T branded x86 system, especially one with a 386, in a single combined package represents the canonical starting point of the UNIX/Linux world as it exists today.

I've done some quick Googling a bit today and found various blog posts, forum posts (including here!) and other resources that detail these AT&T 6386 systems. I'll need to dive in and check these all out. These two links are interesting, but looks like the systems are running DOS instead of UNIX:

AT&T 6386SX/EL WGS
https://www.vogons.org/viewtopic.php?f=46&t=47732

The AT&T PC 6386 WGS
http://0xea.blogspot.com/2013/11/the-at-pc-6386-wgs.html
 
ScutBoy said:
Yes - we ran a bunch of AT&T WGS386's as point-of-sale systems, and we had AT&T branded UNIX media for them that looked just like this photo. I may even have some meda someplace...
Click to expand...

Would you happen to know how these systems were purchased? For example, were they direct ordered from AT&T or via reseller? I'm curious as to the cost and distribution of these systems.

My next foray into this topic will probably include checking out mid-80s magazines to find mentions of either this WGS 386 system and/or direct AT&T releases of SVR4. I'm really curious why this combination of hardware and software was not so popular, particularly in the workstation market. With the release of the 386SX, I would think an AT&T UNIX system should have been able to significantly undercut HP-UX and Sun system costs at the time. But then again, that just may be me looking back at the past with today's lenses and there may not have been any "killer" application for a UNIX x86 system. After all, it is a relatively recent development that *nix systems broke into the mainstream with Android and Mac OS. Not to mention, the Wintel juggernaut would have just been around the corner at this time.
 
atdt916 said:
Would you happen to know how these systems were purchased? For example, were they direct ordered from AT&T or via reseller? I'm curious as to the cost and distribution of these systems.
Click to expand...

Hmm - I'm that's a good question; I know we had a dedicated sales rep for these, but don't know if we he worked directly for AT&T or for a reseller. When I started, the company had 3B1/UNIXPCs in the stores with terminals attached for the point-of-sale, and a 3B2/600 as the back end system at headquarters. These were all upgraded to the 386 systems in the stores, with a dual-486 StarServer/E at the back end. The original 3B* systems and then the x86 stuff all came through the same salesman.

I looked at your links, and the AT&T branded 386 machines we had look different than the two you call out. I'll see if I can find a photo of what we ran. All ours also had QIC tape drives for backups.
 
AT&T was a bit new at the direct sales game in the 80s. The big divestiture free'd them up from an earlier anti-trust consent decree that for-bayed them from selling computers commercially.

All of our 3B2s were actually leased from AT&T and returned after they were taken out of service. It's my understanding this was pretty common among 3B2 installations at least. After most of those machines went off-lease and EoL, AT&T used a lot of them to carpet-bomb universities and sent the others to the land-fill. One of the reasons a series of machines fairly common in the 80s is near extinct today.
 
atdt916 said:
In my opinion, having an AT&T SVR4v2.1 release included with an AT&T branded x86 system, especially one with a 386, in a single combined package represents the canonical starting point of the UNIX/Linux world as it exists today.
Click to expand...

It is much more confusing than that because of the other vendors x86 Unix products including Microsoft/SCO and Interactive Systems' product, which IBM sold.
I was talking to a Unix greybeard on Tuesday, and he reminded me that you can thank Onyx Systems for getting AT&T to create the Unix binary license in 1981.
Before that, you had to spend $25,000 for a Unix source license.

I think the 6386 WGS is an interesting canonical example of Unix before AT&T teamed with Sun (who made their own weird 386i).
There are many paths not taken by Unix clones on the 386, like support for Xenix binaries.
This was also when networking was in flux and AT&T was pushing Starlan when the rest of the
world was going Cheapernet (10-baseT).
 
I seem to recall that before 1981, there were two tiers of Unix source licenses. There was an "educational" license sold to universities and the like and it was nowhere near the $25K for a commercial license. AT&T also briefly toyed with the idea of a lower initial license, but with a per-seat payment. It was a strange time for Ma Bell, culminating in the 1982 consent decree. Another item sometimes missed was that after divestiture, Unix distribution was handled by a now-forgotten spinoff, called AT&T Computer Systems, which, in turn was a spinoff of AT&T Information Systems. The whole affair lasted until about 1991 and was remarkable for its management ineptitude.

Judge Green has since written that he didn't like the consent decree, but by the time it arrived on his desk for signature, it was a done deal and he had no choice but to ratify it. I believe that the Bell System breakup was one of the biggest blunders in US corporate history.
 
I was doing some more digging on the various UNIX versions and I think I found some more answers, but yet another question! :)

Here's an interesting tidbit from Eric S. Raymond's March 31st, 1992 posting on comp.unix.sysv386, in his "SVr4 and clone hardware buyer's FAQ" post. In it, he said:

AT&T's own 386 UNIX offering is not covered here because it is available and supported for AT&T hardware only.
Click to expand...

It seems that AT&T wanted to avoid the competition against Xenix and all the other vendors, and was content on selling vendor licenses for competing in the open-box 386 market. That would explain why their own releases were not as popular as the other vendors, the larger market simply couldn't obtain it without buying AT&T hardware. From the various 80s/90s magazines that I've read on books.google.com, Xenix had the large majority of the x86 UNIX market and was the defacto SVR3 openbox king. With SVR4, AT&T continued to let the vendors fight it out.

Around the time SVR4.2 was being developed, AT&T made a last ditch effort to try their hand in the open-box 386 market, by forming Univel with Novell. Their press release stated:

The companies anticipate that Univel's products will enable computer users to run UNIX applications on standard hardware and utilize the UNIX system as a scalable applications environment across computer networks. Product development is intended to tightly integrate NetWare network services from Novell with the UNIX SVR4 applications platform.
Click to expand...

This effort would produce the first UnixWare release, which is SVR4.2 + NetWare + 'Destiny' desktop stuff. Soon after the release of UnixWare 1.0, AT&T would decide to exit the market and sell USL to Novell, leaving UNIX entirely. That being said, AT&T's last two UNIX efforts were the base SVR4.2 build and the first UnixWare release. That lead me to thinking that these disk images would be from that base SVR4.2 AT&T build, without the UnixWare/Univel additions.

As I was looking further on the possible release date for these disk images, I'd guesstimate that they would have come out during the 1992-1993 timeframe. I found this press release on the SVR4.2 completion date, dated June 16th, 1992:

UNIX System V Release 4.2 will be licensed in source code form by USL to computer industry hardware and software vendors worldwide. Those vendors will provide commercial "binary" versions for their customers.
Click to expand...

AT&T, not being a vendor and the owner of USL, simply produced their own build and released it along with their systems. About six months after that, AT&T and Novell signed their agreement to transfer UNIX System Laboratories from AT&T to Novell. This was finalized on February 16th, 1993:

Tuesday announced the signing of a definitive agreement with AT&T for the acquisition of UNIX System Laboratories, developer of the UNIX operating system.

The definitive agreement has been approved by the boards of directors of both Novell and AT&T, and follows the letter of intent signed by the two companies on Dec. 20, 1992.
Click to expand...

It's not clear that AT&T stopped all development efforts and maintenance of their own internal port at this time (was USL their development and support arm for all things UNIX?), but I'd say it's safe to date these disk images to between June 1992-Feb. 1993 and distributed only with their hardware.

Now for the kicker:

When I first found these disk images, it was from WinWorldPC.com. But now, I found another release on archive.org, which I think is the original unmodified disks that the WinWorldPC.com release was based on. The readme included with the disks from archive.org is dated Dec. 21st, 1999, but the disk image .DCF files themselves are dated January 28th, 1992, which is before SVR4.2 was even completed. So now, that leads me this: Is "Release 4.0 Version 2.1" not the same as "SVR4.2"?!

Apparently not, it seems!

If "Release 4.0 Version 2.1" is not SVR4.2, then likely this is just the last SVR4 release and that the only SVR4.2 build readily available is the UnixWare 1.0 installers that are around the internet. I have not encountered any disk images for Univel's SVR4.2 release that Eric S. Raymond's buyers guide references.
 
alecv said:
It should be possible to use Microport's development tools since both systems are from AT&T

https://archive.org/download/microport
https://archive.org/download/MicroportSystemVDocumentation

(not tested)
Click to expand...

It's a little more complicated than that, since there was divergence in the header files and libraries.
I dumped an Interactive Systems and Bell Technologies V.3 release. The 3.1 floppies I read have the
libraries and header files, now I need to compare them. It would be nice if I can just find the 3.2.3
tools, though.
 
The Microport and AT&T 6386 versions SysVR3 are very close, 3.2.2 (Microport) and 3.2.3 (AT&T 6386).
ISC version is slightly different.

Unfortunately both Microport and AT&T 6386 SysVR3 have no SGS package (C compiler, assembler, headers, libraries e.t.c.)

P.S.
To extract AT&T 6386 SYSV_386_3.2.3_1.44mb_2user:
Code: 
$ ../dumpfloppy/imdcat 323_BASE2.IMD -o 323_BASE2.IMG -n
where 'dumpfloppy' is a Adam Sampson's tool:
https://offog.org/code/dumpfloppy/

Skip 18432 bytes and extrach with `cpio`
Code: 
$ dd if=323_BASE2.IMG of=323_BASE2_.IMG bs=1 skip=18432
$ cpio -itdum -H odc < 323_BASE2_.IMG 
bin
bin/acctcom
bin/ar
bin/basename
bin/cat
bin/clear
bin/cmp
bin/copy
...
 
I've successfully run Microport V/386 (System V 3.2.2) on the PCem v14. The only issue is a video - it does something wrong with VGA so use GCA/80.
The emulated machine is "AMI 486", HDD type is 9 (900/15/17), 1 FDD 1.2Mb.

It does not have development tools (cc compiler, headers e.t.c.) BUT it has restricted development tools from the ID (Installable Drivers) SYSVR3 subsystem.
/lib/idcomp
/lib/idcpp
/bin/idas
/bin/idld
Code: 
# /lib/idcomp -V
C Software Development System 4.1.5 5/27/88
...

AT&T SYSV 3.2.3 seems is fully compatible by _userspace_ and has more recent "idcomp":
Code: 
C Software Development System 4.1.6 9/26/88
It seems another binaries from the AT&T also works on the Microport kernel.
I did not try yet to run AT&T SYSVR3 on any emulators.
Was the AT&T 6386 compatible with conventional PC or not ?

Since there is "as" and "ld", it is possible to build "helloworld" in pure AT&T i386 assembler using direct call to kerlel in "i386 SYSV ABI" style.
Code: 
	.file	"test.s"
	.version	"02.01"
	.set	WRITE,4
	.set	EXIT,1
	.text
	.align	4
	.globl	entry
entry:
	pushl	%ebp
	movl	%esp,%ebp
	subl	$8,%esp

	pushl	$14		/length
	pushl	$hello
	pushl	$1		/STDOUT
	pushl	$0
	movl	$WRITE,%eax
	lcall	$0x07,$0
	addl	$16,%esp

	pushl	$0
	movl	$EXIT,%eax
	lcall	0x07,$0

	.data
	.align	4
hello:
	.byte	0x48,0x65,0x6c,0x6c,0x6f,0x2c, 0x20,0x77,0x6f,0x72
	.byte	0x6c,0x64,0x21,0x0a,0x00
Code: 
# /bin/idas test.s
# /bin/idld -s -e entry -o test test.o
# ./test
Hello, world!
#
 
You must log in or register to reply here.
Back
Top