Ideas versus Implementations

[syzygy]

Similarly, a 10-question medical self-assessment would have copyright on the specific questions asked.

The subject-matter of the questions would not be protected, only their expressions in natural language (and only in so far as that expression is not dictated by the content of the question).

[hyatt]

Similarly, a 10-question medical self-assessment would have copyright on the specific questions asked.

The subject-matter of the questions would not be protected, only their expressions in natural language (and only in so far as that expression is not dictated by the content of the question).

So long as you don't just copy the questions and change words here and there. However, this case is not about a 10-question questionaire, it is more like 50-100 questions. And it is more about the specific aspects of each question. Not a general question, but one with several parts. I think this continual trying to simplify everything into something that can't be copyrighted, by just looking at bits and pieces, is not the way to make progress here. Ditto for the individual eval terms that nobody claims are copyrightable, but when you juxtapose a specific set of eval terms, that are coded in a specific way, the "whole" is copyrightable. And the "whole" clearly violates ICGA rule 2, which is what the ICGA investigation was all about. "Was rybka original or not?" was the question. "no" was the answer...

[syzygy]

So long as you don't just copy the questions and change words here and there. However, this case is not about a 10-question questionaire, it is more like 50-100 questions. And it is more about the specific aspects of each question. Not a general question, but one with several parts. I think this continual trying to simplify everything into something that can't be copyrighted, by just looking at bits and pieces, is not the way to make progress here.

I think your reaction is just funny. I only commented on the example given by Mark.

[BB+]

Does the collection of evaluation features of a chess engine express the personal creative choices of its author? Or did the author only attempt to choose a collection that makes the evaluation function as strong as possible given the rules of chess?

I tend to the latter: creating a good evaluation function is a question of engineering, not of being an artist. Chess optimisation is not fundamentally different from efficiency optimisation.

The Panel essentially found this contention to lack validity. The extent to which strength determines the evaluation function (particularly the choice of features) was concluded to be much less than you indicate. They could be wrong (and some do continue to argue this), but that was the finding, at least for Rule #2 purposes. Indeed, this point has been discussed previously. See RECAP page 19 and page 23 (§4.2.5 and particularly footnote 30), and my Riis rebuttal (page 10). The latter states:

Riis here also seems to suggest that all world-class engines use essentially the same evaluation features. The ICGA Panel considered whether this was likely to be the case, and found it (often from personal experience) to be woefully implausible. As but one example, Rybka 2.3.2a has a Fruit-based evaluation function, while Rybka 3 has one re-written by Larry Kaufman. Both were world-class engines, yet their evaluation functions are markedly different.

More recently, I think I saw that LK stated that he didn't find the [allegedly Fruit-based] R232 eval to be that great (my estimate is the same -- unfortunately, his post seems to have disappeared into TalkChess Engine Origins?!), and I don't see why anyone would particularly think two independent programs would both tend to such a similar suboptimum. In page 19 of the RECAP, I quote VR as posting: There's nothing worse than a big mess of untested eval terms -- if he had actually tested the Fruit evaluators while importing them one-by-one, splicing the most valuable ones into his previous pre-Rybka eval, I would not expect such a notable Rybka/Fruit overlap. Instead, other than the material imbalance table, it's hard to find much of pre-Rybka that remains in the R1 (or R232) eval. [See this post of VR for a vague idea of how eval in pre-Rybka worked].

63. Let us return to the example of the computer program for the reservation of airline tickets. The structure of the program will define the program’s functionalities and describe the combination of those functionalities. The very function of the program, that is to say, to enable the user to obtain an airline ticket, will dictate that combination. It will have to enable the user to check whether the flight exists and, if so, on what date and at what time, whether there are any seats left, and so on. Whatever its nature and scope may be, it is my view that the functionality, or indeed the combination of several functionalities, continues to be comparable to an idea and cannot therefore be protected, as such, by copyright.

I don't like this example, particularly if its scope is made too broad. It's like taking a specific type of book, such as travel guides, noting that plots are (usually) absent, and concluding that plot-copying is irrelevant for almost all literary works. Different genres of computer programs can have differing aspects that come into play. In the case of hand, the very function of the program [that is to say, to play chess as well as possible] is not overly determinative of the evaluation features, at least not in current programs to the extent seen in Rybka/Fruit.

[syzygy]

I tend to the latter: creating a good evaluation function is a question of engineering, not of being an artist. Chess optimisation is not fundamentally different from efficiency optimisation.

The Panel essentially found this contention to lack validity. The extent to which strength determines the evaluation function (particularly the choice of features) was concluded to be much less than you indicate. They could be wrong (and some do continue to argue this), but that was the finding, at least for Rule #2 purposes.

I think the last part is important here: for Rule #2 purposes. Rule 2 doesn't say that it is copyright that matters, and it is obvious that even if copyright would allow precisely reproducing the functionality of an existing engine, Rule 2 still forbids two engines that make the exact same moves.

In the case of hand, the very function of the program [that is to say, to play chess as well as possible] is not overly determinative of the evaluation features, at least not in current programs to the extent seen in Rybka/Fruit.

But the same applies to chess games between humans: the moves are far from perfect(*). What is important is that they were selected in an effort to win the game, not in an effort to give expression to creative freedom.

Two players agreeing to replay a Kasparov-Karpov game are not even playing to win, but do not infringe the copyright on the K-K game, since that game is unprotected (because it was played to win, or at least not to give expression to creative freedom). So as long as Fruit's combination of eval features was created for the purpose of having a good eval and not for expressing Fabien's creative freedom, copying it cannot infringe... (from a copyright point of view).

(*) I will not try to give a definition of "perfect". In a mathematical sense all outcome-preserving moves are perfect, so that is not what I meant.

[BB+]

I've been quite sick for more than a week, but finally have a smidge of time to address some issues a bit more.

Your "idea" of a book is not (only) an idea. You can't abstract your expression of this idea without losing the expression "it was the best of times, it was the worst of times".

The context of this has been lost (it came from 6 months ago) I fear, but I think my abstraction was essentially akin to what Schüle was arguing at the time, and Riis mentions (page 18 of his PDF) in his ChessBase series: But Watkins' evaluation feature is in actuality the formula expressed by an algorithm. This formula is on the conceptual level and therefore, according to accepted practice, everyone is free to use it. Thus his entire analysis lacks traction. I argue that similarly their "formula" is not (only) a formula, but something much more specific than the "conceptual level" they assert.

For instance, I consider something sufficiently broad like "rook on 7th (and 6th/8th)" to be the "concept", and then look at the specific ways that engines express this. As I note in my rebuttal: But even the example given by Riis fails his purpose: some engines give a rook on 7th bonus all the time; some only do so when the opponent's king is on the 8th or an enemy pawn is on the 7th; some add a bonus for doubled 7th rooks (perhaps depending on whether they mutually guard each other), some for a rook and queen both on the 7th; some have a similar bonus for a rook on the 8th (or even 6th), some don't (see §2.3.5 of EVAL_COMP).

The primary purpose of copyright on software is to protect the source code and object code. Everything going further than that might be "nice to have" for whoever profits from the added protection, but it is not essential for the proper functioning of copyright on software. [...] What is important is that copyright protects the final outcome: the code.

I think we diverge at Point Zero. I start from the fact (good or not) that computer programs are protected as literary works. Thus the "primary purpose" of copyright on computer programs is exactly the same as for literary works. To barb your phrase the other way: it might be "nice" to not have to worry about such things as nonliteral copying for computer works, but the law states otherwise. You can argue that (for instance), since the primary purpose of copyright is to "promote the Progress of Science and Useful arts", the ensuing public interest generally (greatly) restricts the protection aspects beyond source/object code, but this is still something that needs to be pled specifically in any given case, rather than just assumed.

So as long as Fruit's combination of eval features was created for the purpose of having a good eval and not for expressing Fabien's creative freedom, copying it cannot infringe... (from a copyright point of view).

I think your argument can similarly be applied to the Rorschach inkblots [they were created for a "functional" purpose, and not as an expression of HR's creative freedom]. Am I missing something?

Rule 2 doesn't say that it is copyright that matters, and it is obvious that even if copyright would allow precisely reproducing the functionality of an existing engine, Rule 2 still forbids two engines that make the exact same moves.

Hmm, I think one of the more recent arguments of Riis was (in essence) that Rule 2 should be taken to mean "original" in the copyright sense (at least for ex post facto application)... About a year ago, Chris Whittington (and maybe Alan Sassler) were arguing similarly regarding the standard in Rule #2. As it was becoming clear that "plagiarism" was going to the the interpretation (rather than copyright), they argued that surely a commercial programmer like VR should be judged by something closer to a copyright standard than some miasmic ivory-tower invention of the ICGA. A related concern, of how to parse words "original" and "derive/derivative", has also been brought up (by Dalke, and MvK too I think with his code/entry distinction). Dalke went so far as to (in private correspondence) suggest that Levy in his interview deliberately intermixed the casual meaning of derivative with the legal definition (though he also seemed to suggest the latter was poorly-defined, and I must say the USC specification is not too illuminative).

[User923005]

In the US, the following is true:

Extent of Copyright Protection
Copyright protection extends to all the copyrightable expression embodied in the computer program. Copyright protection is not available for ideas, program logic, algorithms, systems, methods, concepts, or layouts.

See:
www.copyright.gov/circs/circ61.pdf

Also likely relevant:
http://www.ladas.com/Patents/Computer/C ... t.USA.html
http://en.wikipedia.org/wiki/Software_copyright
http://www.shk-dplc.com/cfo/articles/copyrite.htm

Of course, software laws vary from country to country.

I guess that nobody on the panel has any formal training in software law.
To say that use made of Fruit's ideas was more than fair use may require expert opinion.
Of course, everyone may have their own personal opinion on the subject.
But to make international announcements in the press about a person's guilt should require actual expertize and not some backyard guestimates by untrained people.
IMO-YMMV

[lmader]

But to make international announcements in the press about a person's guilt should require actual expertize and not some backyard guestimates by untrained people.
IMO-YMMV

I have to disagree with this characterization. This was not a court of law, this was a group enforcing their private organization's rules. The standard of proof and for compliance with copyright law was not the same. Also, these were certainly more than "backyard guestimates by untrained people". The people involved were in any case very knowledgeable about software engineering, and an enormous amount of work went into the process. And finally, Vas was given opportunity to provide a defense, which he refused. That left them with little choice.

I guess MMDV (My Mileage Does Vary)

[User923005]

If all they did was strip Vas of his title and publish their result in the ICGA magazine, then I might agree.
But they made international press releases, creating quite a stir.

I agree that the people involved were software experts. But none of them were software law experts.

[Jeremy Bernstein]

If all they did was strip Vas of his title and publish their result in the ICGA magazine, then I might agree.
But they made international press releases, creating quite a stir.

I agree that the people involved were software experts. But none of them were software law experts.

They made international press releases when he won the titles, too.

Jeremy