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?"

Re:Finally...

[MoonFog]

They are far from bankrupt, and probably won't be for a while either. They've already played down their accusations, perhaps trying to have people forget them. Perhaps people will go about their business as they did before this thing started, personally, I hope IBM takes action and drags their sorry faces into the mud.

Wade'da'minute...

[Dark Coder]

The article is a good start but what are the criterias for determining derivatives?

Which method is covered for source code comparisions?

1. two printouts held together and up toward a lighted source?
2. side-by-side subjective eyeball comparision
3. diff (and all derivative comparision tools)
4. diff with some wiggle-room command line options?
5. NSA-grade pattern analysis supercomputer?

I'm slightly guarded here, but these SCO FUD-busting articles seemed very promising...

Re:I found this out a while ago...

[Mr Guy]

So wait, you're trying to say they are EXACTLY THE SAME?

News?

[Anonymous Coward]

This almost doesn't even count as news.

Everybody knew it . . . even SCO!

Re:Finally...

[erick99]

So now, after months and months of news about this trial, it's all over now ?

Since Dr. Randall Davis is an expert witness for IBM, I am guessing that SCO will say, "ain't so!" and then they will ask for time to refute Randall's findings and perhaps come up with an expert witness of their own that finds thousands of "matches." Hopefully the judge in this case will recognize Randall for the expert that he is and accept his findings. However, that just doesn't seem likely to me. This is just another round in a case that will continue like this ad nauseum.

Erick

Counterexample DIY

[hummassa]

1. Get Linux 2.6.8.1
2. Get Linux 2.4.0
3. left out as an exercise for the reader
4. Show positive result
5. Don't profit, but have fun.

Re:$550 an hour....and he reviewed 15 lines of cod

[mekkab]

Great! Graduate Phi Beta Cappa (summa cum laude, too), run some AI centers and also have excessive experience in Code copyright infringement cases!

See you in 10 years!

(trans: read the relevant parts of his CV in the PDF- this guy is FOR REAL.)

Re:At $550 per hour...

[Anonymous Coward]

At $550 per hour, I would've used something like a 386 processor with 8MB of RAM.

I guess that's one reason they didn't hire you.

Re:15 hits

[stratjakt]

to compare all 26,759 lines of the IBM Code identified by SCO against all 67,797,569 lines in the Unix System V Code

All the SCO bullshit over? Far from it. There are still a few hundred million lines of AIX that haven't been compared.

And even if it's over for IBM, doesn't make it necessarily over for Linux in general.

Re:Counter example would have helped.

[mindstrm]

He doesn't need to; the software in question has probably been used many times before for the exact same thing.

Everyone is getting so far off base on this.

SCO is manging to convince people that this is somehow difficult to prove.. that they need more research and more time to PROVE that IBM stole code and put it in linux. Their only claim as to why they think Linux has SCO code is "because there is no way linux could have become as good as it did without stealing from us".. ie: denial

They have yet to show ONE section of code that was lifted. They haven't even shown how one was *similar* enough to have potentially been stolen and heavily modified.. they have shown *NOTHING*

IT's called an expert witness... and their word DOES mean something to the court.. they stake their reputation on it.

Formal Request to Randall Davis

[HopeOS]

Although your deposition includes a description of your methodology, it does not indicate whether you established a proper baseline for comparison or how you calibrated your filter. I would be interested to know how far, in your direct experience, code can be modified before it fails to match COMPARATOR and SIM respectively. Furthermore, how closely does the point at which these tools fail to detect a match coincide with the legal Abstraction, Filtration, and Comparison test?

I do not fault your analysis; I would like to know more about your methodology, beyond the limited scope of the deposition.

-Hope

Re:In other news...

[DarKnyht]

And don't forget Slashdot readers will nitpick.

LAW

[Skiron]

Law is an ass, and there is no way to tell which way this will still come out - on all the arguments.

It only takes a . out of line to sway the legal result, not necessary the correct and right result.

It not over until the fat penguin sings, then we can all rejoice.

Obligatory stock graph

[hackstraw]

Well, it looks like we only have 2 more months [yahoo.com] until SCOX is back to where it should be.

$550 an hour... for having an impressive resume

[Anonymous Coward]

The big bucks are not for what he did, but because his personal history as an expert is sufficintly impressive that judges and juries will find him particularly credible.

Even though what he did may only require a CS degree, if IBM just hired someone with a CS degree to do the same job the SCO lawyers might hire an expert with a better looking resume and be able to convince a non-technical judge/jury that their side was correct.

It's sort of a credentials arms race.

Re:Finally...

[barawn]

Since Dr. Randall Davis is an expert witness for IBM, I am guessing that SCO will say, "ain't so!" and then they will ask for time to refute Randall's findings and perhaps come up with an expert witness of their own that finds thousands of "matches." Hopefully the judge in this case will recognize Randall for the expert that he is and accept his findings. However, that just doesn't seem likely to me. This is just another round in a case that will continue like this ad nauseum.

Dr. Davis is the person who first elucidated how you compare code (the "abstraction, filtration, comparison" test - Computer Associates vs. Aitai) to see if it violates copyright. SCO will have a hard time trying to argue that its depositions (which are from non-experts, though they claim 'unnamed' experts performed the work) are from people more qualified than Dr. Davis.

So I guess what I'm saying is that SCO will have a hard time finding an expert witness more qualified than Dr. Davis. (Please note that if they try to present a deposition from one, that will likely be stricken - as SCO has been ordered by the court to present such a deposition, and has not - thus indicating it doesn't have one) And I highly doubt that the court will value any other expert over Dr. Davis anyway.

SCO has two of its own employees (Dr. Davis is not an IBM employee, though he is being retained by IBM). IBM has the expert witness who first defined how you compare code. Hmm, I wonder which the judge will believe...

Re:Formal Request to Randall Davis

[barawn]

Furthermore, how closely does the point at which these tools fail to detect a match coincide with the legal Abstraction, Filtration, and Comparison test?


Um. Dr. Davis is the guy who first came up with the abstraction, filtration, comparison test - he was the expert witness in Computer Associates vs. Altai. Check his credentials in the first section.

He actually addresses the point you're asking - the code actually finds looser matches than would be found with abstraction, filtration, comparison. So he just ran them through that, said "well, no matches" - since it's a looser comparison, a stricter comparison would be of no benefit.

I think the court will give him the benefit of the doubt that he knows how to do something that he was the first one to do.

Re:lawyers

[nosredna]

If there were such a law, there could be no public defenders or district attorneys. Besides, as cynical as it sounds, the legal system does not really work on the basis of 'truth,' it works on the basis of law. The job of the lawyers is to reveal and even sensationalize as much of the truth that's in their favor as possible, while hiding or downplaying what there is that's against them. It balances out, because he's working against another guy doing exactly the same thing.

Re:Old news, McBride has already admitted this

[rusty0101]

If you paid attention to the FA, you might have noticed that what McBride is accusing IBM or other Linux developers of doing, includes changing the variable names to hide the 'stolen' code, and that the software that the Dr. used takes this possibility into consideration, and would find such lines if they existed.

The basic result is that no such lines existed that can be demonstrated to be non-literal copying, or literal copying.

-Rusty

Hardly a surprise but there's more

[UnknowingFool]

While news of no infringing code is hardly a surprise to everyone here, it really explains a lot of SCO's behavior. Everyone believes there probably isn't literal copying in Linux. SCO probaly does too. But SCO is stuck having made so many public statements. What they've been trying to do is sneak in the derivatives, modifications, "methods and concepts" copyright infringements. This explains their steadfast and unbending request for all of AIX and Dynix source code. They know that nobody copied SysV and put it into Linux. But many people at IBM (and others) may have used SysV as a guide to develop their own code. They want to get their hands on anything that looks like somebody at IBM wrote that might be similiar enough to pass off as having some lineage to SysV. Problem for SCO is that current copyright laws do not support this notion.

Re:Counter example would have helped.

[darkonc]

I'm sure they will jump on the fact that the expert didn't provide an example of a true derivative work run through the same procedure.

It just hit me: He doesn't have to. It's SCO's responsibility to show that there is infringing code in Linux. It's not IBM's responsibility to show that there is none. All that Davis has to prove is that the search is feasible in a reasonable ammount of time (as opposed to SCO's claim of 25,000 man-years). He's done this admirably. Not being able to find anything is simply icing on the cake.

One beautiful thing about this is that (AFAICT) all (or almost all) of the software he used seems to be Open source (although he has references some similar commercial software), so SCO has absolutely no excuse to not repeat his experiment and come up with different results (presuming that they've actually got a case), given that it takes about 1 hour to run the comparison on off-the-shelf hardware.

The other beautiful thing about this is -- remember Darl's remarks about an MIT team deep-diving the code?...... (boot to the head!)
"I've shown you mine, now you show me yours!"

Re:Counter example would have helped.

[theMightyE]

SCO is manging to convince people that this is somehow difficult to prove... They have yet to show ONE section of code that was lifted. They haven't even shown how one was *similar* enough to have potentially been stolen and heavily modified.. they have shown *NOTHING*

I've been thinking this was strange too. After all, if code was copied into Linux it is essentially a public document now - out there for everyone to see. All SCO would have to do is download it and print it out side-by-side with a copy of their matching code. Case closed. SCO wins.

The fact that they haven't done this extremely simple thing seems to strongly point to SCO being a bunch of total bullshitters. Even if some malicious programmer intentionally stole code and modified it slightly (changed variable names, comments, re-arranged the order of functions in header files, etc.) it should be pretty trivial to show a judge what happened and move on to the 'get sacks of cash from IBM' phase of the trial.

Funny, you would think that a company that is suffering continuous, ongoing harm to the tune of US$699 per user would be pretty quick to do such a thing...

Re: a judge will weigh.

[mink]

I thought Gupta did not qualify as an expert (IBM pointed this out) so his thing is only what he thinks not matters of fact.

Justice upgrade

[Doc Ruby]

If US justice required evidence to be registered and certified allowable before claims were filed, the system would shed a lot of its unbearable load. Even the appeals before a judge, of disallowed evidence, would be dealt with in a more efficient manner. It's insane that I'm paying for the legal system that SCO is exploiting to promote its equity, now that its legitimate business has failed, while their travesty hasn't even got any evidence, or demonstrated any basis for their claims. After the years they've pushed this thing through the courts, they should have at least produced some evidence. And they're just the flagship corporate operation lawsuit - billions are spent by my cohort of taxpayers to keep litigious corporations and their lawyers in business. We should nip them in the bud, by simply requiring evidence to make any claims based on it.

Re:Finally...

[Zordak]

In a court case like this, the two sides are expected to retain and pay expert witnesses to advise the court. The experts are expected to be honest because their reputation, which is worth quite a lot (in this case $550/hr) is riding on their testimony. In fact, at Wednesday's hearing, Judge Kimball basically asked SCO why they hadn't retained any experts to counter IBM's numerous real experts. SCO lamely responded that it was still too early (15 months into the case). What they didn't (and couldn't) say is that they haven't retained any experts because no expert is going to sign his reputable name to their idiot theories. The only testimony that SCO has offered is that of Chris Sontag and Sandeep Gupta, two SCO employees. They were paid too. The difference is, Dr. Davis is able to submit a formidable resume that qualifies him to offer expert testimony. Sontag actually tries to sound qualified by saying that he has an MIS degree and took some computer classes in college. Gupta has an engineering degree. The problem is, SCO can't even decide what kind of testimony these guys are giving. It was supposed to be personal knowledge (which anyone can give regarding what they have seen directly), but they then went on to give what amounted to expert testimony (which is supposed to be based on deep, extensive knowledge in the field, like Dr. Davis'). Then SCO's lawyers went on to tell the court that, althogh Sontag and Gupta gave personal experience testiomny, they qualified as experts, and their testimony was relevant, but Dr. Davis' wasn't.

The point is, this is not IBM paying some front man to do their dirty FUD work and dress it up as an "independent" study. Everybody in the court knows that IBM is paying Dr. Davis as they are expected to. What is telling is that SCO has been unable to offer any counter-testimony from any real experts. They've still got $50 million in the bank. It's not because they can't afford it.

Re:Thanks Professor Davis... and thanks ESR...

[fiftyfly]

I hope that you got the patches for the supposed "major performance bottlenecks" that Mr. Davis mentioned. It's seems a little funny to me that Mr. Davis made $550 an hour for the difficult job of running SysV and Linux through comparator and commenting on the output. At those rates you would think he would be grateful for performance bottlenecks.

He didn't get $550/hr to run comparator, he got the fee for being an expert recognizable as such to the court and damned near irrefutable on the subject. His reputation earned him $550/hr.

Re:Sounds like....

[surprise_audit]

Has anyone during this entire process given thought that maybe SCO took the code and claims it as their own? How will they ever prove this case is beyond me, since Unix was in existence before SCO was established as a company.

Actually, didn't Prof Davis also just prove that SCO's source doesn't include Linux code?? If SCO had stolen anything and included it in their code, it would have shown up in the comparator test, wouldn't it?? The comparison just shows common code, it doesn't distinguish which is the original and which is the copy.