First, public access to "unpublished" copyright files is generally (but not completely) restricted.

Second, even if you do get access, the code in the file isn't likely to help in many cases. As a copyright registrant you need only usually provide the first and last 25 pages of code (or changes), with variations on that rule for machine code, trade secrets (which can be blacked out), and so on... this is in the computer code copyright circular that can be found on the copyright office web site. It also means that the copyright registration file for computer code is a difficult source for evidentiary support.

It makes the task you are suggesting (while important) daunting and practically impossible, and raises one of the major unanswered questions in modern copyright law as it applies to computer code -- how to prove exactly what is a derivative work in computer-code-land (I'm talking about "proof" on paper for copyright registration, not so much the abstraction-filtration-expression test).

Otherwise, though, thanks for the great weblog.

[ Reply to This | # ]

I can't comment on the details, but certainly SCO's copyright claims remain nebulous - especially the parts about "methods and concepts" which are not covered by copyright at all. Claims like that reinforce the impression that SCO are targeting ill-informed managers who just see the words "alledged IP infringement" and panic.

I very much doubt that if IBM registered the JFS code, it gave the copyright to SCO ;)

Also note that although SCO alledges further that System V code was copied into the kernel verbatim by "unix vendors", it conveniently omits to say who, even though this could be easily established from the mailing list archives. If you don't name anyone, they can't sue you for libel ;) However, we do know one unix vendor who openly contributed code to the kernel as part of company policy - Caldera!

Brillig's point is well-made: but it cuts both ways for SCO. In the case of allegedly copied code, the burden of proof is on them to show they wrote it first. Given that timestamps on files can be effortlessly forged, a lack of signed, dated, paper printouts acts against them. Remember that the analysts shown the famous "80 lines" noted that all dates had been removed from both listings - why? Surely SCO would leave the dates in if "their" code had an earlier date?

SCO have not even contacted Red Hat - is this the action of a company protecting its IP? They made 2.4.13 source code available for months, without any change to its stated license or copyright notice, from their FTP site (and apparently it's still available from a few urls), even though dozens of people publicly expressed amazement this was happening - would a company which honestly believed it owned IP therein do this? ftp.kernel.org has received no legal notice from SCO telling it to stop offering the kernel source code - again, what *sane* management would allow this blatant infringement of IP to continue? Are SCO intent on setting a legal record by allowing every possible affirmative defence by IBM to apply many times over?

[ Reply to This | # ]

[ Reply to This | # ]

Therefore, the only JFS technology that *anyone* could have copyrighted in 1992 was JFS1, but that was entirely in IBM's hands - so the documents filed in 1992 couldn't cover that. Even if they did, they wouldn't cover JFS for Linux which is not related to the JFS1 code.

Moreover, unless I'm very much mistaken, System V *to this day* has neither JFS1,JFS2, or JFS. SCO's claims over JFS reside entirely on their contractual argument that because IBM created JFS1 for a SystemV derivative os (AIX), all other implementations of JFS - even clean-room - are theirs. If this sounds risible, it's probably because it is.

I'm not sure about RCU & NUMA, but the story is probably similar. For example: SCO's own products don't support NUMA. Sequent wrote NUMA and SCO was allowed use of that code in Project Monterey. SCO now seems to think this means they own it. Maybe there was a contractual agreement in Monterey not to re-use the code outside the project, but that's a contractual dispute, not copyright.

SMP is a red herring. Alan Cox wrote the original support himself (with Caldera's blessing!) and it would be impossible, given the different kernel architectures, for his work to infringe on System V's copyright. For SMP to scale well beyond 4 cpus, you need things like NUMA and RCU, which is why SCO drags them in. But the original SMP support was in 2.2, and SCO have indicated the 2.2 kernel is ok, so by their own mouths they have blown that claim out of the water.

Which brings us back to the beginning - SCO is claiming contractual violations by IBM (and Sequent, now part of IBM) relating to JFS, RCU and NUMA. The JFS aspect is not only baseless but irrelevant since Linux already had other journalling filesystems available to make it enterprise ready, and SCO's product doesn't have JFS. That leaves RCU and NUMA. Even if IBM violated the Monterey contract in giving them to Linux, that would not give SCO copyright ownership. And even if some weird interpretation of the original AT&T contract gave SCO copyright ownership of the RCU and NUMA code, both segments of code could be easily removed from the kernel and re-written, since SCO holds patents for neither. Therefore SCO's refusal to send a letter to Linus saying "please remove the RCU and NUMA code" nets them "forty laches" (pardon the pun)

[ Reply to This | # ]