Randall Davis: IBM Has No SCO Code

Mick Ohrberg writes "As reported by Groklaw, Randall Davis, renowned professor of Computer Science at MIT has after an extensive search found no evidence of SCO's claims that IBM has incorporated parts of the Unix System V code. Davis says "Accordingly, the IBM Code cannot be said, in my opinion, to be a modification or a derivative work based on the Unix System V Code." Surprised, anyone?"

Comments

Thanks Professor Davis... and thanks ESR...

[LinuxParanoid]

... who wrote the comparator tool which was one of the two tools used in the analysis.

ESR deserves three cheers for 'scratching his itch', making a tool to compare copyrighted code. To have it actually used in the SCO case which was the annoying impetus for its creation (AFAICT) has to be a nice feeling.

I'm not an ESR fanboy, but I'll give him props when I think he deserves it and in this case I think he does.

--LP

Read the PDF...

[mekkab]

He goes into detail.

a judge will weigh.

[www.sorehands.com]

Judges will weigh the 'testimony' of experts in a case. In some cases, a court will appoint an expert as a special master -- to make technical determinations.

If I recall correctly, Randy told me that he has served as a special master in several cases.

Re: a judge will weigh.

[shotfeel]

Though this is on one level a matter of expert opinion, and on another level a matter of fact.

One would hope, as a matter of fact, the SCOexpert would be required to show where he found matches. That (if it exists) can be explicitly shown to the court.

Then, if need be, the experts can argue over wether or not they match.

Kind of like fingerprints. The suspect's fingerprints are entered as evidence, as are fingerprints found at the scene. The experts can then argue about wether or not they match. But until those fingerprints are presented and accepted as evidence, there is no weighing of testimony to be done.

Re:question

[Carnildo]

Novell. There's a lawsuit going on about the matter right now, but it looks like the Unix copyrights were never actually transfered from Novell to SCO.

Re:Finally...

[Nurseman]

I hope IBM takes action and drags their sorry faces into the mud.

This has been gone over at length on Groklaw. IBM HAS taken action. No matter what SCO does, IBM still has a huge countersuit under something called Lanham Act . Methinks SCO is in a bit of trouble

Re:At $550 per hour...

[Anonymous Coward]

He's not charging for CPU time, he's charging for his own time. The Xeon timing bit was included, I assume, to demonstrate that the tests were not prohibitively difficult. This may be intended to preemptively counter a SCO claim that testing for this was impractically difficult.

I expect it took more than 10 hours for him to write that document and painstakingly verify its accuracy and wording to avoid perjury. He'll likely spend significant amounts of time testifying in person on the subject as well.

That rate is high but within reason for top-end expert witnesses (which is exactly what he is.) It's not uncommon for renowned professors to make a substantial second income by acting as an expert witness (very common in the chemistry and biology fields, at least.)

Finally, IBM would not even blink if they were handed a bill for several hundred hours at $550 each on this issue. They may even get some of the money back, depending on the details of the final settlement and the subsequent SCO bankruptcy.

Re:question

[MathFox]

With Novell holding Unix copyrights

That raises an interesting possibility for Novell. I wonder if they've thought about integrating their Unix copyrights into their Linux distribution and later suing other makers (and, heck, users) of Linux distributions for copyright infringement.

The moment Novell releases Unix code as part of Linux the Unix code will be distributable under the GPL. The Unix code has become part of Linux and the only way to distribute the whole is under the GPL. As copyright owner Novell will still be able to dual licence the same code to Sun, MS, etc.

Citing the GPL:

If identifiable sections of that work are not derived from the Program, and can be reasonably considered independent and separate works in themselves, then this License, and its terms, do not apply to those sections when you distribute them as separate works. But when you distribute the same sections as part of a whole which is a work based on the Program, the distribution of the whole must be on the terms of this License, whose permissions for other licensees extend to the entire whole, and thus to each and every part regardless of who wrote it.

Clear enough?

Heh, paragraph 30

[xant]

(copied by hand, not pasted, since the PDF is of a fax of a copy or somethign..)

The box below shows one of the reported matches from the lines of code cited by SCO. COMPARATOR reported a match between lines 588-591 in rclock.c and lines 1665-1667 from System V UW1.1 /src/i386at/uts/io/target/sdi.c:

Lines 588-591 from rclock.c
#endif /* RCLOCK_PROF */
return;
}

Lines 1665-1667 from sdi.c
#endif
return;
}

The two "words"--endif and return--that appear in the two files are so common in code written in the C language that finding them together like this is purely an accident, of no significance in detecting copying. In particular, the code from each file above simply signifies the ending of a routine; it is as if we had found two bodies of unrelated English text that each happened to conclude with the words "the end".

I think that pretty much sums up this whole case from the beginning.

Re:What did Didio see?

[dgatwood]

Simple. She saw similar code. Look, Laura DiDio is a tech analyst. She looks for trends in the market. She is neither a programmer nor a copyright expert. Davis is both. It's easy for SCO to pull one over on someone who doesn't have the skill set needed to understand what does and doesn't constitute infringement, but far harder to do so to someone who specializes in the subject. :-)

Davis saw no -infringing- code. That's not the same thing as seeing no common code. Copying a ten-line function almost verbatim is likely a copyright violation. Ending up with a handful of lines that look a lot alike, by contrast, is often just an unavoidable side-effect of writing two pieces of code that do the same thing.

Small blocks of code (under 3 lines) are generally not protected by copyright (unless we're talking obfuscated C lines). Even larger blocks of code may not be protected, depending on content.

For example, in many cases (drivers come to mind), there may only be exactly one way to do something (e.g. you must set this register to 1, wait 5 ms, set it to zero, wait 5 ms, then set a second register to 1), in which case those specific bits are unlikely to be copyrightable at all, even if they represent a fairly substantial number of lines of code.

Also, in order for the code to be infringing, it must have been taken from AT&T UNIX or its descendants, and must not have been put there by someone who owned copyright on said code. That means that A. the code must not have been placed there by anyone working for SCO or Novell, and B. the code must not have come from a third, shared source such as BSD. A very large chunk of SCO's UNIX code fails the "B." test, and SCO was an active contributor to Linux, so many cases where code appears the same could easily fail the "A." test as well.