Fruit-Rybka and Copyrights.

[Chan Rasjid]

Hello,

This again re-visits the same issues.

Vasik cloned Fruit. He went through the codes of Fruit "forward and backward and took many things" and, as the codes of Fruit is simple and clear, understood almost all of it. By implementing it in bitboard and, maybe adding some very good original ideas of his own, had Rybka which is stronger than all the other programs including the then strongest engine Shredder. As Bob Hyatt found out, Vasik probably also did some "cut-and-paste" of codes from Crafty to his earliest Rybka versions. So that was said to be the genesis of the earliest Rybka.

What about copyrights? Did Vasik Rajlich infringe on the copyrights of Fruit. I know nothing about copyright laws except what I gathered and learned from what others posted in threads here. So I don't pretend anything. But the basic idea of copyrights seems simple if we use some common sense. The basic principle is to protect intellectual work of publishing authors. The content of what is published may be irrelevant. It could be a sillest of novel, my implementation of bubble sort, the first 10,000 digits of the mathematical constant pi - in other words, anything published. The basic of infringement should also be simple - it is just whether someone copied from the work of another. But there cannot be any non-trivial copyrights law if it is only about making exact copies. So copyrights extends to derivative work when it could be ascertain that the a work could be traced to have been derived and based on another's work that is protected by copyrights.

So did Rybka infringe on the copyrights of Fruit. I think very likely as the findings by the by the ICGA panel showed - that there were too much similarity. Similarity means copying and derivation from the source of Fruit. So Rybka infringes on the copyrights of Fruit. This is my (new ?) view and it is also the view of many others. But of course the ultimate test is with the courts.

Now going to the issue of PST. In copyrights, I think it is irrelevant where the bishop's table contain a lot of original information or whether most rely on something that can be encapsulted in a simple array of {-3, -1, 0, 1,...} that nearly all of us use as the same starting point. Again copyrights is simply just about "copying". If the FSF were to bring the Fruit-Rybka case to court, the test probably would be about copying and derivation. The issue of whether PST by nature has enough of complexity probably would not be taken into consideration. If the balance of probability is that the work could not have been done except through copying or derivation, then there is infringement of copyrights. So just the bishop's PST alone matching may not mean much, but all 12 x 64 values matching could give rise to copyrights issue unless there is clear proof they are arrived at independently.

So what should we do about implementing PST in our program. It should not be a problem. Any aspiring chess programmer can just plug in the exact formulae from Fruit (or Robbolit, Stockfish) to start with. It is simple practicality. PST is the last thing on the todo list and why should someone waste time on tuning something that is "least" important and when there are so many other things to tune. The Robbo values, if it makes a 3000ELO program, cannot be too wrong for anyone. So cloning here completely solves a problem. I started off with (probably) the exact formulae of Fruit's PST and never lost a wink of sleep. No one would ever sue me or, in the worst (or best) case scenario, I could call up the help of random numbers and use static tables.

Now what about cloning the search of another program? Again for "original" programs, I think it should never be a problem. The conditions for search are very few and it is never a big problem, say, to clone that of Stockfish; of lmr, futility pruning, SEE, nullmove, etc. My own program is in C and I think the basic data structure are so different that even if I meticulously clone the search parameters of Stockfish, it could never be detected in a manner to have an issue of copyrights infringement. There are not many ways to write an alpha-beta search and to set search parameters and conditions. The methods are already well known. Also comparison of binaries would not be able to bring out any blocks of similar code except through the strangest of concidence. So what Bob Hyatt said that if we started out with an "original" program, then very likely we are safe, is correct. Other little embedded cloning would then become normal and acceptable just as the ocean, because of the it's original content and vastness, could never be found polluted.

I write this post as, for some reason or other, I have a change in perception about the Fruit-Rybka cloning and copyrights infringement issue. I did made some (many) posts in the past expressing my views on the issue and about the acceptability of cloning in general. This would (probably) be the last post that I would make concerning cloning. With this change of perception, my other recent thread that I started would be irrelevant and it would be better if I see that the moderators remove it.

Best Regards,
Rasjid.

[Michel]

This again re-visits the same issues.

Why?

[Chan Rasjid]

This again re-visits the same issues.

Why?

I don't know what you mean. Maybe there is subtlety that is beyond me.

Rasjid.

[Dirt]

It could be a sillest of novel, my implementation of bubble sort, the first 10,000 digits of the mathematical constant pi...

No, someone could reasonably come up with the same 10,000 digits without ever seeing yours. You may have put in a lot of effort to come up with those digits, but there is too little creativity involved.

[michiguel]

Hello,

Now going to the issue of PST. In copyrights, I think it is irrelevant where the bishop's table contain a lot of original information or whether most rely on something that can be encapsulted in a simple array of {-3, -1, 0, 1,...} that nearly all of us use as the same starting point. Again copyrights is simply just about "copying". If the FSF were to bring the Fruit-Rybka case to court, the test probably would be about copying and derivation. The issue of whether PST by nature has enough of complexity probably would not be taken into consideration.

It should.
You cannot copyright concepts, and even less the ones that are known, line "centralize" and "punish edges".

If the balance of probability is that the work could not have been done except through copying or derivation, then there is infringement of copyrights. So just the bishop's PST alone matching may not mean much, but all 12 x 64 values matching could give rise to copyrights issue unless there is clear proof they are arrived at independently.

There are not 12x64 values matching. This is misinformation that has been spread.

- There are 11 tables.
- 4 of them are just plain different
- 7 are "matches" (one trivial), which means, they can be obtain with the mechanics used in Fruit (but the numbers itself are different and do not respect the proportionality). As I showed in R forum, they could also be obtained with different mechanics based on first principles.
- If VR "copied" the PST, it means he copied a handful of vectors like
[-3,-1,0,1] that contain the concept "centralize" and "punish edges". All the alleged multipliers and parameters to obtain the tables would be different.

To illustrate how simple this is, Queen tables and K endgame would also be a match for B tables!

The information content here is extremely low, and based on rudimentary well know chess knowledge. This is illustrated by the fact that one "match" is also found in Crafty (with a complete different mechanics as it was shown) and many in Stockfish (from Glaurung, of course).

There were extensive threads in R and Open chess forum, with different opinions, of course. After dedicating quite a bit of time on this, I am fully convinced that the evidence on PST should not have been included as incriminating because this cannot be "copyrighted". It only suggests that VR studied Fruit, which should be necessarily bad. In fact, it sets a dangerous precedent.

I do not have intentions to rehash this all over again, so please read the R forum and open chess.

Miguel

So what should we do about implementing PST in our program. It should not be a problem. Any aspiring chess programmer can just plug in the exact formulae from Fruit (or Robbolit, Stockfish) to start with. It is simple practicality. PST is the last thing on the todo list and why should someone waste time on tuning something that is "least" important and when there are so many other things to tune. The Robbo values, if it makes a 3000ELO program, cannot be too wrong for anyone. So cloning here completely solves a problem. I started off with (probably) the exact formulae of Fruit's PST and never lost a wink of sleep. No one would ever sue me or, in the worst (or best) case scenario, I could call up the help of random numbers and use static tables.

Now what about cloning the search of another program? Again for "original" programs, I think it should never be a problem. The conditions for search are very few and it is never a big problem, say, to clone that of Stockfish; of lmr, futility pruning, SEE, nullmove, etc. My own program is in C and I think the basic data structure are so different that even if I meticulously clone the search parameters of Stockfish, it could never be detected in a manner to have an issue of copyrights infringement. There are not many ways to write an alpha-beta search and to set search parameters and conditions. The methods are already well known. Also comparison of binaries would not be able to bring out any blocks of similar code except through the strangest of concidence. So what Bob Hyatt said that if we started out with an "original" program, then very likely we are safe, is correct. Other little embedded cloning would then become normal and acceptable just as the ocean, because of the it's original content and vastness, could never be found polluted.

I write this post as, for some reason or other, I have a change in perception about the Fruit-Rybka cloning and copyrights infringement issue. I did made some (many) posts in the past expressing my views on the issue and about the acceptability of cloning in general. This would (probably) be the last post that I would make concerning cloning. With this change of perception, my other recent thread that I started would be irrelevant and it would be better if I see that the moderators remove it.

Best Regards,
Rasjid.

[michiguel]

Hello,

Now going to the issue of PST. In copyrights, I think it is irrelevant where the bishop's table contain a lot of original information or whether most rely on something that can be encapsulted in a simple array of {-3, -1, 0, 1,...} that nearly all of us use as the same starting point. Again copyrights is simply just about "copying". If the FSF were to bring the Fruit-Rybka case to court, the test probably would be about copying and derivation. The issue of whether PST by nature has enough of complexity probably would not be taken into consideration.

It should.
You cannot copyright concepts, and even less the ones that are known, line "centralize" and "punish edges".

If the balance of probability is that the work could not have been done except through copying or derivation, then there is infringement of copyrights. So just the bishop's PST alone matching may not mean much, but all 12 x 64 values matching could give rise to copyrights issue unless there is clear proof they are arrived at independently.

There are not 12x64 values matching. This is misinformation that has been spread.

- There are 11 tables.
- 4 of them are just plain different
- 7 are "matches" (one trivial), which means, they can be obtain with the mechanics used in Fruit (but the numbers itself are different and do not respect the proportionality). As I showed in R forum, they could also be obtained with different mechanics based on first principles.
- If VR "copied" the PST, it means he copied a handful of vectors like
[-3,-1,0,1] that contain the concept "centralize" and "punish edges". All the alleged multipliers and parameters to obtain the tables would be different.

To illustrate how simple this is, Queen tables and K endgame would also be a match for B tables!

The information content here is extremely low, and based on rudimentary well know chess knowledge. This is illustrated by the fact that one "match" is also found in Crafty (with a complete different mechanics as it was shown) and many in Stockfish (from Glaurung, of course).

There were extensive threads in R and Open chess forum, with different opinions, of course. After dedicating quite a bit of time on this, I am fully convinced that the evidence on PST should not have been included as incriminating because this cannot be "copyrighted". It only suggests that VR studied Fruit, which should be necessarily bad. In fact, it sets a dangerous precedent.

I meant... "which should NOT be necessarily bad"
Miguel

I do not have intentions to rehash this all over again, so please read the R forum and open chess.

Miguel

So what should we do about implementing PST in our program. It should not be a problem. Any aspiring chess programmer can just plug in the exact formulae from Fruit (or Robbolit, Stockfish) to start with. It is simple practicality. PST is the last thing on the todo list and why should someone waste time on tuning something that is "least" important and when there are so many other things to tune. The Robbo values, if it makes a 3000ELO program, cannot be too wrong for anyone. So cloning here completely solves a problem. I started off with (probably) the exact formulae of Fruit's PST and never lost a wink of sleep. No one would ever sue me or, in the worst (or best) case scenario, I could call up the help of random numbers and use static tables.

Now what about cloning the search of another program? Again for "original" programs, I think it should never be a problem. The conditions for search are very few and it is never a big problem, say, to clone that of Stockfish; of lmr, futility pruning, SEE, nullmove, etc. My own program is in C and I think the basic data structure are so different that even if I meticulously clone the search parameters of Stockfish, it could never be detected in a manner to have an issue of copyrights infringement. There are not many ways to write an alpha-beta search and to set search parameters and conditions. The methods are already well known. Also comparison of binaries would not be able to bring out any blocks of similar code except through the strangest of concidence. So what Bob Hyatt said that if we started out with an "original" program, then very likely we are safe, is correct. Other little embedded cloning would then become normal and acceptable just as the ocean, because of the it's original content and vastness, could never be found polluted.

I write this post as, for some reason or other, I have a change in perception about the Fruit-Rybka cloning and copyrights infringement issue. I did made some (many) posts in the past expressing my views on the issue and about the acceptability of cloning in general. This would (probably) be the last post that I would make concerning cloning. With this change of perception, my other recent thread that I started would be irrelevant and it would be better if I see that the moderators remove it.

Best Regards,
Rasjid.

[Chan Rasjid]

It could be a sillest of novel, my implementation of bubble sort, the first 10,000 digits of the mathematical constant pi...

No, someone could reasonably come up with the same 10,000 digits without ever seeing yours. You may have put in a lot of effort to come up with those digits, but there is too little creativity involved.

I am not sure you understood me. I am interpreting copyrights from first principle and not the real thing. I would take copyright infringement to mean there is an "act of copying". I think if Vasik could use an equivalent defense of alibi, "my eyes did not ever visit the codes of Fruit", then he would be clearly innocent.

Maybe in copyrights law, there may be things that cannot be copyrighted, as you seem to suggest, eg. triviality or commonly known facts.

Best Regards,
Rasjid.

[jdart]

Hello,

Now what about cloning the search of another program? Again for "original" programs, I think it should never be a problem. The conditions for search are very few and it is never a big problem, say, to clone that of Stockfish; of lmr, futility pruning, SEE, nullmove, etc. My own program is in C and I think the basic data structure are so different that even if I meticulously clone the search parameters of Stockfish, it could never be detected in a manner to have an issue of copyrights infringement. There are not many ways to write an alpha-beta search and to set search parameters and conditions. The methods are already well known. Also comparison of binaries would not be able to bring out any blocks of similar code except through the strangest of concidence. So what Bob Hyatt said that if we started out with an "original" program, then very likely we are safe, is correct. Other little embedded cloning would then become normal and acceptable just as the ocean, because of the it's original content and vastness, could never be found polluted.

It sounds like you are mixing up several questions. The first is, is a certain kind of copying illegal (violates copyright)? The second question is, is it acceptable (morally)? The third question is, is it detectable in a binary?

Re the first question, we're not copyright lawyers and what's legal and illegal is being litigated frequently. So you can make some assumptions but the lines are not always clear and there can be different legal theories and opinions.

In general algorithms are not subject to copyright. Can you copyright a numeric table? I don't actually know .. It may depend on how much creativity went into designing it. Generally in my experience code that contains tables of constants as key ingredients (such as some crypto algorithms) is protected in other ways than copyright, such as a patent.

I'm not going into the 2nd question. Re the 3rd question, don't underestimate what a good disassembler can do, although it is very laborious to disassemble a program.

[Chan Rasjid]

Hello,

Now going to the issue of PST. In copyrights, I think it is irrelevant where the bishop's table contain a lot of original information or whether most rely on something that can be encapsulted in a simple array of {-3, -1, 0, 1,...} that nearly all of us use as the same starting point. Again copyrights is simply just about "copying". If the FSF were to bring the Fruit-Rybka case to court, the test probably would be about copying and derivation. The issue of whether PST by nature has enough of complexity probably would not be taken into consideration.

It should.
You cannot copyright concepts, and even less the ones that are known, line "centralize" and "punish edges".

If the balance of probability is that the work could not have been done except through copying or derivation, then there is infringement of copyrights. So just the bishop's PST alone matching may not mean much, but all 12 x 64 values matching could give rise to copyrights issue unless there is clear proof they are arrived at independently.

There are not 12x64 values matching. This is misinformation that has been spread.

- There are 11 tables.
- 4 of them are just plain different
- 7 are "matches" (one trivial), which means, they can be obtain with the mechanics used in Fruit (but the numbers itself are different and do not respect the proportionality). As I showed in R forum, they could also be obtained with different mechanics based on first principles.
- If VR "copied" the PST, it means he copied a handful of vectors like
[-3,-1,0,1] that contain the concept "centralize" and "punish edges". All the alleged multipliers and parameters to obtain the tables would be different.

To illustrate how simple this is, Queen tables and K endgame would also be a match for B tables!

The information content here is extremely low, and based on rudimentary well know chess knowledge. This is illustrated by the fact that one "match" is also found in Crafty (with a complete different mechanics as it was shown) and many in Stockfish (from Glaurung, of course).

There were extensive threads in R and Open chess forum, with different opinions, of course. After dedicating quite a bit of time on this, I am fully convinced that the evidence on PST should not have been included as incriminating because this cannot be "copyrighted". It only suggests that VR studied Fruit, which should be necessarily bad. In fact, it sets a dangerous precedent.

I meant... "which should NOT be necessarily bad"
Miguel

I do not have intentions to rehash this all over again, so please read the R forum and open chess.

Miguel

So what should we do about implementing PST in our program. It should not be a problem. Any aspiring chess programmer can just plug in the exact formulae from Fruit (or Robbolit, Stockfish) to start with. It is simple practicality. PST is the last thing on the todo list and why should someone waste time on tuning something that is "least" important and when there are so many other things to tune. The Robbo values, if it makes a 3000ELO program, cannot be too wrong for anyone. So cloning here completely solves a problem. I started off with (probably) the exact formulae of Fruit's PST and never lost a wink of sleep. No one would ever sue me or, in the worst (or best) case scenario, I could call up the help of random numbers and use static tables.

Now what about cloning the search of another program? Again for "original" programs, I think it should never be a problem. The conditions for search are very few and it is never a big problem, say, to clone that of Stockfish; of lmr, futility pruning, SEE, nullmove, etc. My own program is in C and I think the basic data structure are so different that even if I meticulously clone the search parameters of Stockfish, it could never be detected in a manner to have an issue of copyrights infringement. There are not many ways to write an alpha-beta search and to set search parameters and conditions. The methods are already well known. Also comparison of binaries would not be able to bring out any blocks of similar code except through the strangest of concidence. So what Bob Hyatt said that if we started out with an "original" program, then very likely we are safe, is correct. Other little embedded cloning would then become normal and acceptable just as the ocean, because of the it's original content and vastness, could never be found polluted.

I write this post as, for some reason or other, I have a change in perception about the Fruit-Rybka cloning and copyrights infringement issue. I did made some (many) posts in the past expressing my views on the issue and about the acceptability of cloning in general. This would (probably) be the last post that I would make concerning cloning. With this change of perception, my other recent thread that I started would be irrelevant and it would be better if I see that the moderators remove it.

Best Regards,
Rasjid.

As you say, we do not wish to delve to much into these PST thing here; furthermore I don't have the expertise. But I could say these. I could just revisit the source code of Fruit with the PST and ,just maybe in 10 mins, I could memorise everything and reproduce them from memory. So I don't understand why such things in Fruit should prevent me from using the exact tables as the reasons for the arrays and muliplier values are quite easy to grasp.

I am not sure you can easily make a case that PST should not be used as incriminating evidence. If Vasik did as I did when I used the exact PST stuff, then there would be total exact PST matches. And if what count is "copying is copying", then bad for Vasik - It is is simply the way how copyrights work, not whether we agee or don't agree. The exception is when we could argue on the ground that certain things are not copyrightable as you seem to suggest.

Best Regards,
Rasjid.

[michiguel]

Hello,

Now going to the issue of PST. In copyrights, I think it is irrelevant where the bishop's table contain a lot of original information or whether most rely on something that can be encapsulted in a simple array of {-3, -1, 0, 1,...} that nearly all of us use as the same starting point. Again copyrights is simply just about "copying". If the FSF were to bring the Fruit-Rybka case to court, the test probably would be about copying and derivation. The issue of whether PST by nature has enough of complexity probably would not be taken into consideration.

It should.
You cannot copyright concepts, and even less the ones that are known, line "centralize" and "punish edges".

If the balance of probability is that the work could not have been done except through copying or derivation, then there is infringement of copyrights. So just the bishop's PST alone matching may not mean much, but all 12 x 64 values matching could give rise to copyrights issue unless there is clear proof they are arrived at independently.

There are not 12x64 values matching. This is misinformation that has been spread.

- There are 11 tables.
- 4 of them are just plain different
- 7 are "matches" (one trivial), which means, they can be obtain with the mechanics used in Fruit (but the numbers itself are different and do not respect the proportionality). As I showed in R forum, they could also be obtained with different mechanics based on first principles.
- If VR "copied" the PST, it means he copied a handful of vectors like
[-3,-1,0,1] that contain the concept "centralize" and "punish edges". All the alleged multipliers and parameters to obtain the tables would be different.

To illustrate how simple this is, Queen tables and K endgame would also be a match for B tables!

The information content here is extremely low, and based on rudimentary well know chess knowledge. This is illustrated by the fact that one "match" is also found in Crafty (with a complete different mechanics as it was shown) and many in Stockfish (from Glaurung, of course).

There were extensive threads in R and Open chess forum, with different opinions, of course. After dedicating quite a bit of time on this, I am fully convinced that the evidence on PST should not have been included as incriminating because this cannot be "copyrighted". It only suggests that VR studied Fruit, which should be necessarily bad. In fact, it sets a dangerous precedent.

I meant... "which should NOT be necessarily bad"
Miguel

I do not have intentions to rehash this all over again, so please read the R forum and open chess.

Miguel

So what should we do about implementing PST in our program. It should not be a problem. Any aspiring chess programmer can just plug in the exact formulae from Fruit (or Robbolit, Stockfish) to start with. It is simple practicality. PST is the last thing on the todo list and why should someone waste time on tuning something that is "least" important and when there are so many other things to tune. The Robbo values, if it makes a 3000ELO program, cannot be too wrong for anyone. So cloning here completely solves a problem. I started off with (probably) the exact formulae of Fruit's PST and never lost a wink of sleep. No one would ever sue me or, in the worst (or best) case scenario, I could call up the help of random numbers and use static tables.

Now what about cloning the search of another program? Again for "original" programs, I think it should never be a problem. The conditions for search are very few and it is never a big problem, say, to clone that of Stockfish; of lmr, futility pruning, SEE, nullmove, etc. My own program is in C and I think the basic data structure are so different that even if I meticulously clone the search parameters of Stockfish, it could never be detected in a manner to have an issue of copyrights infringement. There are not many ways to write an alpha-beta search and to set search parameters and conditions. The methods are already well known. Also comparison of binaries would not be able to bring out any blocks of similar code except through the strangest of concidence. So what Bob Hyatt said that if we started out with an "original" program, then very likely we are safe, is correct. Other little embedded cloning would then become normal and acceptable just as the ocean, because of the it's original content and vastness, could never be found polluted.

I write this post as, for some reason or other, I have a change in perception about the Fruit-Rybka cloning and copyrights infringement issue. I did made some (many) posts in the past expressing my views on the issue and about the acceptability of cloning in general. This would (probably) be the last post that I would make concerning cloning. With this change of perception, my other recent thread that I started would be irrelevant and it would be better if I see that the moderators remove it.

Best Regards,
Rasjid.

As you say, we do not wish to delve to much into these PST thing here; furthermore I don't have the expertise. But I could say these. I could just revisit the source code of Fruit with the PST and ,just maybe in 10 mins, I could memorise everything and reproduce them from memory. So I don't understand why such things in Fruit should prevent me from using the exact tables as the reasons for the arrays and muliplier values are quite easy to grasp.

I am not sure you can easily make a case that PST should not be used as incriminating evidence. If Vasik did as I did when I used the exact PST stuff, then there would be total exact PST matches. And if what count is "copying is copying", then bad for Vasik - It is is simply the way how copyrights work, not whether we agee or don't agree. The exception is when we could argue on the ground that certain things are not copyrightable as you seem to suggest.

Best Regards,
Rasjid.

But VR did NOT "copy" the tables. It is suggested and inferred that may have copied the mechanics of how they were built and a handful of simple structural vectors. You cannot copyright these simple loop mechanics. The discussion should be if those vectors could be copyrighted. I personally think: no.

What you say is a bit different but it is in line with Richard Vida. He says that PST should not even be considered an issue.

You say "just maybe in 10 mins, I could memorise everything and reproduce them from memory". Correct. In fact, when I looked at them, I realized the tables could be derived from first principles rather than vectors. That is because the vectors imply simple first principles!

These are important discussions that need to be taken place. If this evidence is used against R1, there are many engines that automatically become "suspect" even when no code has been shared.

Miguel