Before I passed the reins of Groklaw content to Mark Webbink, I had obtained some additional early AT&T/Regents of California documents from the University of California. Thanks to Mark, I finally had time to scan them in and upload them and add them to Groklaw's permanent collection. I'd like to share some highlights with you, at his request.
What stands out to me the most is that AT&T licensed BSD code from the Regents, repeatedly and very early. I will show you the agreement. And another document confirms that AT&T didn't register copyrights on 32V, relying on trade secret protection only, back when you had to take certain steps to own valid copyrights.
Novell, now Attachmate, may think it owns all the pre-1995 UNIX copyrights, thanks to Novell's victory over SCO at trial, but all that was established at trial was that Novell did not *transfer* any pre-1995 copyrights it owned to SCO. The trial didn't parse out and establish what precisely Novell owned, what it had obtained from AT&T in the first place, or what AT&T owned, for that matter.
This is for historians, mainly, but should the SCO v. IBM case ever ramp up again or any new litigation occur, it could come in handy to have these documents.
Here they are:
- Western Electric/Regents License Dated Dec.1, 1973
- Western Electric Letter to Regents 1978
- AT&T Letter to Regents April 5, 1982
- AT&T Letter to Regents Sept. 15, 1983
- Regents Letter to AT&T Dec. 17, 1984
- Regents Letter to AT&T Jan 30, 1985
- AT&T Letter to Regents Feb. 12, 1985
- AT&T-Regents Educational License 1985
- AT&T Letter Agreement to Regents Jan. 28, 1986
- AT&T Letter to Regents Feb. 26, 1986
- Regents to AT&T License Agreement, March 4, 1986
- University of California April 1, 1986 4.3 BSD
- AT&T 1988 Announces CPIO Source Code
- University of California Jan. 1989 Re First Networking 4 BSD
- AT&T Letter to Regents April 24, 1989
- Keith Bostic 1993 Memo Re Last Release of 4.4 BSD
- CSRG Archive CDs BSDs 1999
Let's go in more or less chronological order. It is, of course, fascinating to see the 1973 (actually signed in 1974) license agreement [PDF] between Western Electric and Regents. It was a license to use the software to teach students at the University of California, under confidentiality terms, but I see that it included this phrasing:
... however, such obligations shall not extend to any information or technical data relating to the LICENSED SOFTWARE which is now available to the general public or which later becomes available to the general public by acts not attributable to LICENSEE, its employees or students. IBM had such language in its agreement in 1985 with AT&T if you recall. It's the home-free card, I'd say, given the widespread public knowledge that has by now taken place.
In that 1973 agreement, there was no allowance to license to others at all or do derivatives, but that changed pretty soon, as we can see in the 1978 letter [PDF] from Western Electric to Regents:
In response to your letter of January 18, 1978 to Mr. A. L. Arms, we take no position with regard to your use or distribution of software developed by you which does not contain any of our proprietary information such as, without limitation, the computer programs and documentation, or any portions thereof, related to the UNIX Operating System, including methods and concepts utilized therein. So, at first, it was kind of loose, with restrictions on who could have the materials, yes, but with Western Electric saying it didn't wish to have to do the monitoring in advance of distribution. And significantly, you can see that whatever code the licensee wrote was his, just so long as it didn't include any of Western Electric's proprietary materials. Their homegrown code was theirs, not Western Electric's.
With regard to the distribution of any software which does contain some of our proprietary information, our primary concern is the prevention of unauthorized disclosure, either intentional or inadvertent, which might jeopardize our valuable proprietary rights in such information. We have no objection to your distribution of such software to other licensees having licenses for the same UNIX Operating System software.... The manner of such distribution is within your discretion. However, the distribution of any such software must include a requirement that the recipient treat Western's included proprietary software in the same manner as if such included proprietary software had been received directly pursuant to the license agreement which such recipient has for the UNIX Operating System. We do not wish to assume the burden of verifying the status of potential recipients of such software prior to your distribution thereof.
I see in the AT&T letter of April 5, 1982 [PDF] to the Regents of the University of California that from 1978 to this letter four years later, AT&T's policy was *not* to closely monitor who received BSD as long as it went to those already licensed for UNIX. This letter states that it was altering the policy going forward with respect to UNIX System III:
It is our policy to permit our software licensees to distribute their own versions of source code for a licensed software package to others licensed by us for the same package. Procedures for such distributions were discussed generally in my April 10, 1978 letter to your Ms. Susan L. Graham. You have now requested that you be permitted to distribute your version of source code for 32V to our licensees for UNIX System III> So here we see BSD mentioned specifically. There was no objection to its distribution to other licensees, no objection to any licensees distributing "their own versions of source code for a licensed software package" to others with a license to the same package, meaning they are describing derivative works. There follows a list of how the Regents were to keep records going *forward*, by getting a copy of the licensee's software agreement for System III and then calling AT&T to confirm the status and keep records and send them every three months to AT&T. Four years is a long time not to monitor your trade secrets, if you care.
In my April 10, 1978 letter I stated that we did not wish to assume the burden of verifying the status of potential recipients of such software distributions. We have changed that policy with respect to distributions to licensees of UNIX System III. Accordingly, we propose that you be permitted to distribute your version of source code for 32V to our licensees for UNIX System III provided that you: ...
I read all this to say that it was, for a time at least, quite a bit looser than SCO tried to tell the court in the litigation, where it asserted that AT&T had always scrupulously and without fail maintained strict confidentiality and insisted its licensees do the same.
The following year, in 1983, UNIX System V licensees were
added [PDF] to the list of folks the Regents could distribute BSD to under the same new terms as System III.
However, in 1984, there began a dispute [PDF]. Regents was not happy with what it viewed as new terms presented in an AT&T letter dated July 30, 1984:
This is in response to the above-referenced letter in which you attempt to impose certain conditions on the University of California's licensing of the Fourth Berkeley Software Distribution. You state as follows:
This Regents letter [PDF] from Roy Towers to AT&T's David Frasure dated January 30, 1985 adds to the complaint:
"This is to inform you that Licensees desiring to obtain the Berkeley 4.1 and/or 4.2 distribution of the UNIX Operating System must have a source license with AT&T for the DEC family of UNIX Software, i.e., UNIX 32V, System III, System V, System V Release 1.0, or System V Release 2.0. If a customer has a license from another System V family (AT&T 3B, M68000, etc.), that customer cannot be licensed by Berkeley until licensed by AT&T for the DEC family of UNIX System V."
The University of California cannot adhere to these terms as they represent a unilaterally imposed amendment to our UNIX/32V license agreement, executed by the University on 2 October 1981 and by AT&T on 27 October 1981. Any such restrictions to our distribution of 4.1 or 4.2 BSD would require a formal agreement signed by authorized representatives of both AT&T and The Regents of The University of California.
We don't have the referenced July 30, 1984 AT&T letter, but we do have
this AT&T letter [PDF] of February 12, 1985, where AT&T quotes from the 1984 AT&T letter and responds to the Regents' concerns: "We do not agree with your characterization of such letter as a unilateral amendment," AT&T's D.W. Frasure writes to Roy Towers, at the University of California, referring to a 1981 agreement. AT&T's position in 1985 was that permissions granted in a Sept. 15, 1983 letter (and I would assume in the 1981 agreement) "does not extend to our licensees of versions of UNIX System V not intended for use on DEC computers." It states that the "original versions of UNIX System V were for Digital Equipment Corporation (DEC) machines" and that it never gave Regents permission to distribute BSD to AT&Ts licensees "for the versions of UNIX System V intended for AT&T's 3B computers and computers based on Motorola's M6800 product."
There seems to be some confusion about the above-referenced letter, and I should like to offer some clarification.
Attached please find the following documents:
Correspondence from E.G. Baldwin of Western Electric to Susan L. Graham of the University of California dated 10 April 1978 These letters contain specific instructions for the distribution of Fourth Berkeley Software, and represent a contract between The Regents of the University of California and AT&T. The new restrictions you seek to impose in your letter to Pauline Schwartz of 30 July 1984 represent a unilaterally imposed amendment to this contract that is unacceptable to the University of California.
Correspondence from E.G. Baldwin of AT&T to The Regents of the University of California dated 5 April 1982
Correspondence from D.P. Wilson of Western Electric to The Regents of the University of California dated 15 September 1983
That is how Mr. Frasure read the above April 5, 1982 and Sept. 15, 1983 letter agreements. I see nothing in them matching his interpretation, personally, and the 1978 Western Electric letter seems to contradict his interpretation, and please notice this language on page 1:
The 1979 and 1981 agreements grant to the University of California certain limited rights to use LICENSED SOFTWARE. There are specific provisions in the agreements (i) prohibiting the use of LICENSED SOFTWARE by or for any third person (last sentence of Section 2.01 in both agreements), (ii) requiring that LICENSED SOFTWARE be held in confidence (Sections 5.06 in the 1981 agreement and Section 4.06 in the 1979 agreement) and requiring that LICENSED SOFTWARE not be sold, leased, or otherwise transferred or disposed of (Section 5.10 in the 1981 agreement and Section 4.10 in the 1979 agreement). Notwithstanding these provisions, it has been our policy for a number of years to permit our licensees for a particular software product to furnish their derivative versions of such software products to our other licensees for the same software product. Such policy was not reflected in the language of the software agreements we used in 1979 and 1981, but our software agreements now cover this point. (See Section 7.06(b) in the attached specimen copy of our current Educational Software Agreement. Of course, you were operating under that policy when you distributed your version of 32V to our other licensees for 32V. Emphasis is mine.
Otis Wilson told the truth, then, in his depositions, and as testified to by Dr. Peter Salus in his Declaration [PDF] in SCO v. IBM, that there were policies that changed over time, and sometimes the policies were not reflected in the license agreements. And the licensees were free to do their own versions, without any claims of ownership of their code by AT&T.
Anyway, the point is that in these newly available documents, what we are seeing is AT&T trying to alter a policy midstream, but with Regents not agreeing to the change. I can see why Regents would view the DEC limitation as a unilateral amendment, frankly, absent any other written agreements or letters. But we don't have the full record.
In any case, in a letter dated 1986 it looks like AT&T prevailed, because the letter announces BSD 4.3, and it says it was available to "VAX users with UNIX/32V, System III, or System V source licenses with AT&T", VAX being a DEC trademark.
Then, in this 1989 letter [PDF], the announcement was made by the University that no AT&T license was required at all for the first release of BSD networking software:
We are happy to send you information about our first release of the BSD networking software. This software is copyright Berkeley, and may be freely redistributed. It is available to anyone and requires no previous license, either from AT&T or the Regents of the University of California. You had to pay $400 and sign the attached license agreement. After that, you could "reproduce or distribute any copies of the LICENSED MATERIAL or derivative works, in source or binary form, as long as the original copyright notice is retained by LICENSEE in redistributed source code, and appropriate credit is given to the Electrical Engineering and Computer Science Department of the University of California at Berkeley." They provided wording for credit. But the licensee was free to "enhance, modify, or correct" the code, merge it into other material and then distribute the modified version. That would include AT&T, of course. They could do all those things as well. And in due time, so could Linux.
One thing I've learned from researching UNIX is that bookkeeping wasn't complete back then, and it's worse now, making trying to trace it all out quite difficult, but at least we can fill in some gaps now. SCO's attorney, Kevin McBride -- Darl's brother --
said in open court in 2003 that there are no trade secrets any more in UNIX ("There is no trade secret in Unix system files. That is on the record. No
problem with that."), so you could argue it's all moot. (Again, see Groklaw's
Unix Books project for multitudinous examples of the "secrets" being scattered to the winds long ago, even if the attorney had not said what he said.] But given SCO's tactics, and all the asset sales, it never hurts to have it all carefully documented to the degree we can.
AT&T Licensed *From* the Regents:
Significantly, we have now a March 1986 licensing agreement, signed by AT&T and the Regents of the University of California. But this one isn't Regents licensing from AT&T. It's the other way around. It's AT&T licensing BSD *from* Regents. And by its wording we learn there were earlier versions of the same type. Here it is [PDF]. Please note on page 1 that AT&T acknowledges three significant things:
1) that Regents owned "enhancements and additions to 32V, which together with parts of 32V comprise computer programs and documentation entitled Fourth Berkeley Distribution"; That would be you guys, I presume, some of you, all the students and others who contributed patches and improvements back then.
2) AT&T wanted to take a license from Regents to use BSD 4.2 and 4.3, and that it had earlier, in 1983, licensed 4.1 BSD; and
3) that 4.2 BSD and 4.3 BSD "include some material contributed by persons other than agents, officers, and employees of the University" who are termed in the agreement "Other Contributors" which it says "are identified within 4.2 BSD and 4.3 BSD.
AT&T under this license got the right to use the software and to sublicense it with or without modification whether or not in conjunction with any other software. That means to me that AT&T could incorporate BSD code into its own UNIX software. And we know it did.
On page 2, notice that paragraph 4 says that AT&T agreed that 4.2 and 4.3 BSD contained proprietary software belonging to the University, and AT&T had no "right, title or interest" in that software, other than the rights set forth in the agreement. If you look at paragraph 5, the University agrees that 4.2 BSD and 4.3 BSD "contain proprietary software belonging to AT&T and licensed by AT&T as 32V".
Paragraph 7 says that AT&T got the right to sublicense object code "that an end user accepts by opening the package containing the object code." It could be modified and then sublicensed by the AT&T sublicensee.
So, so far, we see enough to realize that both AT&T's UNIX and the Regents' BSD had code in it owned by the other. So in any litigation, one would want to know precisely who owned the file that might be introduced into the case, along with proof of copyright ownership.
On the next page, you see in paragraph 8, that AT&T was to give "appropriate credit" to the University of California "and to the Other Contributors for their roles in the development and will require sublicensees to give such credit," and they could put it in documentation similar to 4.2 and 4.3 BSD notices. It provides the language:
This software and documentation is based in part on the Fourth Berkeley Software Distribution under license from The Regents of the University of California. We acknowledge the following individuals and institutions for their role in its development: [insert names of individuals and institutions which appear in the documentation provided to AT&T as part of 4.2 BSD and 4.3 BSD for those portions of said Distribution used by AT&T.]" Did AT&T abide by this obligation? We found out in the BSDi litigation that they did not always do so. I wonder if that means the license no longer covers them? You can see that the University could terminate the license on written notice to AT&T if AT&T breached and did not cure. If terminated, AT&T was to destroy all BSD copies it had.
But one thing is obvious -- if anyone, SCO or anyone, tried to sue over UNIX code, they ought to be required to demonstrate in court that they actually owned the code, that it did *not* come from BSD. Given their failure to abide by the credit requirement, that will be at a minimum really hard and at worst impossible. Or at best, depending on where you are standing.
Let's look at the Schedules, beginning on page 6. One page seems to be missing, so if anyone ever finds it, please provide it to us. But what we have is fascinating, because look at what it was that AT&T licensed *from* -- not *to* -- Regents, on page 9 of the PDF, a list of files licensed, including an old friend:
I notice a couple of things:
1) The full schedule isn't there. There is mention of "pages 2-5" but
we only have 2 & 3, and it's clear from the alphabetic listing that
page 3 isn't the last page.
2) The list of files is a *complete* list of all the files on an old
BSD system. In other words, the tape was a dump of a live system. For
example, one of the files is /etc/hosts, which is a configuration file
and something you wouldn't license to anyone.
Now, as you know, the first errno.h
came from 32/V and had about 30 or so errors listed in it. So the fact that this is listed on this as being licensed doesn't mean AT&T didn't own the original. But Berkeley
then added another 30-40 errors for the networking and other
things to the file. In listing the files to be licensed, AT&T
evidently simply wanted to list all the files from which they wished to
take substantial Berkeley code. The listing of a file there
did not imply that it was entirely Berkeley code. Just that
it had substantial Berkeley additions.
But stop and think. SCO sued IBM over errno.h, as if it owned it all. It turned out it didn't own it at all anyway, but even if it had, at most only about half of the errors were AT&T code, and the rest were not. That's consistent with the
1993 complaint in the BSDi litigation, where the assertion was that about half of the code at issue was actually from BSD. BSD was copyrighted. But what about the AT&T-written code?
AT&T wasn't good at keeping the BSD copyright notices on the code, despite being required by the license to do so, so by the time SCO showed up, they might have seen only an AT&T copyright notice and just assumed they owned it, had they owned it. For the sake of historians, though, you can find here the 1983 and 1985 versions of errno.h released with 4.2 BSD. And here's a more modern list.
I contacted Warren Toomey about all this, to make sure of the details, and he told me the following:
BSD certainly implemented networking, and AT&T took a lot of the code
from around the 4.2BSD epoch and used it in SysVR4. This code would have used
many new error numbers which are network-specific. In
you will see the 4.2BSD errno.h; most of the error numbers from 35 up
are network-related. Here's where you can go to compare files between various versions of Unix and non-Unix systems, by the way.
Of course there are many other significant problems for anyone trying to sue over errno.h, which Groklaw has chronicled over the years, including Caldera
releasing Ancient Unix under a BSD-like license. And then there's POSIX compliance. And a significant piece is that AT&T didn't register any code for copyrights in the early days. That came out in the BSDi litigation, if you recall. In Judge Dickinson Debevoise's
ruling on summary judgment, he wrote this:
AT&T's Bell Laboratories
developed UNIX in the 1970s, and registered the name UNIX as
Trademark No. 1,392,203 on May 6, 1986. (1st Am. Compl. Ex. B.)
In addition, AT&T has received copyright certificates of
registration on UNIX software and documentation for the Fifth
through Seventh Editions and for version 32V. (Id., Exs. C-F.).... We see confirmation in AT&T's 1988 announcement [PDF] about copyrighting CPIO Source Code:
In order to prevail, Plaintiff must prove that it has a
valid copyright in the UNIX code. Plaintiff's chief difficulty
here is the "Publication doctrine." The publication doctrine
denies copyright protection to works which the copyright owner
"publishes," unless the owner has properly affixed a notice of
copyright to the published work. This doctrine has suffered steady
erosion over the years, and it now applies in full force only for
works published prior to January 1, 1978. For works such as 32V
(published in 1978), which were published after that date but
before March 1, 1989, the doctrine is subject to the escape
provisions of 17 U.S.C. Section 405(a) and the common-law "limited
publication rule." For works published after March 1, 1989, the
publication doctrine has been eliminated by the Berne Convention
Implementation Act, 102 Stat. 2857 (1988).
Because Plaintiff copyrighted 32V in 1992, Plaintiff
benefits from a statutory presumption of validity and BSDI has the
burden of proving otherwise. Williams Electronics, Inc. v. Arctic
International, Inc., 685 F.2d 870, 873 (3rd Cir. 1982). BSDI seeks
to carry this burden by proving that AT&T and Plaintiff have widely
published 32V without proper notice....
Version 32V source code has now been distributed,
without notice, to literally thousands of licensees.
Consequently, Plaintiff can have no valid copyright on 32V unless
it can fit within one of the statutory or common law escape
The three statutory escape provisions are listed in
section 405(a). These provisions relieve a copyright owner from
the harsh consequences of noticeless publication if the owner
(i) omitted the notice from a "relatively small number of copies;"
(ii) registers the work within five years of publication, and then
makes a "reasonable effort" to add notices to the noticeless copies
already distributed; or (iii) proves that a third party omitted,
notice in violation of an express agreement in writing 17 U.S.C Section
Plaintiff cannot avail itself of any of these provisions.
AT&T has decided to publish, subject to suitable copyright restrictions, a portion of AT&T's source code known as CPIO software and described in Appendix A attached to this letter agreement.... Here we see AT&T acknowledging that it relied on trade secret protection prior to this moment, and that all it was copyrighting was the CPIO software.
Now, consider that the license from Regents to AT&T was in 1986, and adding in the information that AT&T's errno.h code consists only of the first 34 or so files, that 32V was first published in 1978, and then noting that AT&T didn't register a copyright for 32V until 1992, and this judge earlier wrote that it wasn't valid anyhow and that AT&T had no copyright due to lack of proper notices on 32V, how in the world could any entity justify a claim of copyright infringement on the early
UNIX was published without notice, AT&T relying on trade secret protection instead, at a time when that mattered. At a minimum, any defendant would seem to have a right to ask the plaintiff to demonstrate an actual ownership trail and prove the validity of any asserted copyright. And from the 1986 license alone, any plaintiff would have to prove that whatever they are suing over
*didn't* come from BSD 4.3, or 4.2 or 4.1, all of which AT&T licensed
from Regents, going back to 1983, if not before.
Berkeley understands that the decision by AT&T to so publish the CPIO software in no way affects the trade secret status of any other AT&T source code licensed by AT&T to Berkeley or Berkeley's corresponding obligations of confidentiality with respect to such other source code.
So Novell, now Attachmate, may think it owns all the UNIX copyrights after Novell's victory over SCO at trial, but in reality
it owns, at most, a subset and not the basic, early
UNIX code or methods and concepts (in a copyright context). AT&T made a decision to rely on trade secrets instead, and later owners of Unix have to live with its choice.
A couple of other side points: If you review SCO's answers to IBM
interrogatories [PDF] early in the case, where SCO explained its theory
of how IBM allegedly misused "its" code, on pages 21 and 22, you see errno.h in the section
about JFS. SCO's claim was this:
In addition, IBM's drop of JFS into Linux reference files contains
references to UNIX-based header files, not otherwise found in Linux
prior to IBM's identified transfers, further indicating that the source
of this technology was AIX. A reference to sys/errno.h doesn't prove it came from AIX. Any
software developed on a POSIX compliant system (or one with a
partially compliant POSIX implementation) might reference that file.
For example, here is someone developing on OS/2 (the source of Linux
JFS) having to reference that very file.
And while SCO repeatedly claimed that similarities between source files "proved" copying, Groklaw's Dr Stupid explains here how code could be similar without it implying copying, that some similarities are inevitable. I had to put it in a separate page, to keep the code he uses to demonstrate his points intact.
You may notice more highlights in this collection. If so, be sure to leave a comment.
I've put them all on Groklaw's permanent Contracts page, under both BSDi and AT&T headers, so you and historians can always find them easily in any research projects you may undertake.