On Tue, Apr 8, 2025 at 4:25 PM segaloco via TUHS <tuhs@tuhs.org> wrote:

So System V shops had to hold a license with AT&T to modify and redistribute
code based on UNIX System V and they would then license directly with their
customers correct?

You are talking about two different licenses (with three fees, actually).

This being distinct from the way licensing with BSD was
concerned in that you had to pursue the license with AT&T to then use BSD. That
is my current understanding anyway that I base this question on.

Right... that would worked for any version of the UNIX license from Fifth Edition through SVR5, which I believe was the last license that Novell released to the world.

If you are a commercial customer, the license grants you the right to have the sources and use them on a >>single<< CPU, which is named (via make and serial number) in the license. Each additional CPU must be named and licensed separately at a reduced fee.

The universities and research institutions had a non-commercial license that did not name a CPU. They could use the AT&T IP anywhere as long as it was for non-commercial purposes [more in a minute—there is a catch-22].

Either of these licenses also gives you the right to share your sources with anyone else who has a source license from AT&T for the same (or a lesser) version. That is, if you had a System 3.0 [renamed Sys III] or later System V license, you could still obtain things from someone with a V6 or V7 license (like UCB say), but since you were Sytem III and they V7 or V32, you could only give them code that was based on the license they had. For example, SCCS was part of PWB 2.0 originally and officially left the labs as part of and Sytem 3.0. Remember the research license only included V5, V6, V7, and V32 — not PWB, nor System 3 While SCCS "leaked" to some of the Universities via AT&T Employee's (the OYOCs), the Universities were technically not licensed for it, which is why, of course, even though it is being used at UCB, it's not in the BSD source distribution and also why UCB switched to RCS as soon as it was available (because it done at another University and was >>freely<< open source, not just AT&T licensed and open with the AT&T license community.

The interesting catch-22 came about when Universities (like CMU originally) started using UNIX commercially internally for things like billing and student records. A few of us grew uncomfortable about this pretty blatant misuse in Mellon Institute, CMU's semi-cokmerical arm, which was using UNIX as its development system for the projects they contracted (like the Pittsburgh Press automation, USS Steel, Alcoa, and even the Israeli Army. In fact, in mid-1978, Danny Klein and I went on strike at Mellon Institute, and Al Arm's of AT&T Patent and Licensing team offered a compromise where CMU bought one commercial license and named the Mellon Institute PDP-11 as the CMU commercial host. Officially, any work, like the work that started to be done for the CMU back-office, even if it ran on a different system, was officially designated as work for Mellon. I know Case-Western got a similar license in late 1979 because I told Fred Park about it when he was working with us at Tektronix in the Summer of 1979. I believe UCB and Purdue also followed suit, but I never knew.

So IBM, DEC, Sun, HP, Microsoft, etc. approach AT&T, got a source license,

err.. Sun was not there, thank you very much... [they would not exist for another 6 years].

With the release of the V7 license, a group of Prof. Dennis Allison's clients were brought together at Ricki's Hyatt in Palo Alto [I was there, and it is, as I have said in other messages, the only time I ever met Gates]. The firms I remember that were there (besides Al Arms for AT&T) were IBM, Tektronix, DEC, HP, Microsoft, 3Com, and what would eventually become SCO. There were a couple of others, but I do not remember who they were. Somebody like Bruce Borden, who was representing 3Com, might, and Dennis might remember who he invited from his client list.

At that point, AT&T had produced the V7 "redistribution" license, which once you had a source license, then allowed you (for a per CPU fee) the right redistribute the AT&T IP. Let's just say the terms were pretty bad. I'll not go into them here.

This first meeting resulted in a set of follow-up meetings to create what would eventually become the Sytem 3.0 redistribution license (and the new Sytem 3.0 source license). This group would also eventually create what would become /usr/group, but that's another story. As I said, AT&T Marketing decided to change the marketing materials to System III at the last minute. But all the docs and the licenses say System 3.0 because they were already through the print cycle for the former and the legal approval cycle for the latter.

The key is that a commercial firm had to pay for the right to have the source only on >>specific<< CPUs and then the rights redistribution, and then on a per redistributed binary, a license for each of those CPUs (3 different fees). BTW, the whole X-user license stuff for end-user binaries was because of the redistribution license fees. The fee the licensee paid to AT&T for a binary that ran a "workstation" was significantly cheaper than one that ran an eight or 16-user, much less 32 or 64-user time-sharing machine [thank you, Microsoft and DEC actually -- I can explain how this all happened in another email if folks are interested]. With later redistribution licenses for later AT&T releases such as System V, SRV2, SRV3, SRV4, some of the source license storage requirements were relaxed, but for this list, please understand there is the core source license (which is based on each release) and then a redistribution license for the release, and per CPU license

BTW: the primary reason why OSF was created was because everytime AT&T made a new OS release, they changed the redistribution license terms. If you look at the OSF found principles the first one is "stable license terms." It really was a problem for everyone.