Open source or not? Opinions wanted

[Sune Fischer]

Geschrieben von:/Posted by: Sune Fischer at 31 July 2004 22:58:26:
Als Antwort auf:/In reply to: Re: Open source or not? Opinions wanted geschrieben von:/posted by: Tord Romstad at 31 July 2004 15:16:35:

I prefer to view the field more like a science, and if I could do
something that could help *all* engines improve by 20 Elo points it would make me
much more happy than if I managed to improve only my own engine by 200 Elo points.
Dann is right that calculus would have been invented even without Newton and
Leibniz, but I think that progress for the field as a whole is always much faster when
people are willing to share their ideas.

I think that offering to open source the program is very generous, but I'm not sure if that's the best way to go if you're thinking about doing it 'for the good of science'.
Take Crafty for example, a very strong open source program, but kinda hard to read and figure out what's actually going on. Well, in my opinion anyway.
If I had my choice I'd rather read about the good stuff in a more compact and direct form, such as in an article. It's generally a bit easier than reverse engineering somebody elses code.
I know that you have a lot of interesting ideas about how to use the eval to guide the search and improve its stability.
You deserve to become the author of a few popular and frequently quoted articles!
Changing the subject, what's that I read about you removing your attacktables?
-S.

[Uri Blass]

Geschrieben von:/Posted by: Uri Blass at 31 July 2004 23:30:51:
Als Antwort auf:/In reply to: Re: Open source or not? Opinions wanted geschrieben von:/posted by: Sune Fischer at 31 July 2004 22:58:26:

I prefer to view the field more like a science, and if I could do
something that could help *all* engines improve by 20 Elo points it would make me
much more happy than if I managed to improve only my own engine by 200 Elo points.
Dann is right that calculus would have been invented even without Newton and
Leibniz, but I think that progress for the field as a whole is always much faster when
people are willing to share their ideas.

I think that offering to open source the program is very generous, but I'm not sure if that's the best way to go if you're thinking about doing it 'for the good of science'.
Take Crafty for example, a very strong open source program, but kinda hard to read and figure out what's actually going on. Well, in my opinion anyway.
If I had my choice I'd rather read about the good stuff in a more compact and direct form, such as in an article. It's generally a bit easier than reverse engineering somebody elses code.
I know that you have a lot of interesting ideas about how to use the eval to guide the search and improve its stability.
You deserve to become the author of a few popular and frequently quoted articles!
Changing the subject, what's that I read about you removing your attacktables?
-S.

The problem with articles is that part of the process is not about good algorithm but about writing a good code without bugs.
I understand that the main advantage of the new engine of Tord relative to Gothmog is that Gothmog is badly written.
My main problem in improving Movei is not lack of ideas but code that is badly written.
Uri

[Harald Luessen]

Geschrieben von:/Posted by: Harald Luessen at 01 August 2004 00:29:31:
Als Antwort auf:/In reply to: Open source or not? Opinions wanted geschrieben von:/posted by: Tord Romstad at 30 July 2004 16:14:59:

Hi

Right now, I am considering whether I should release the source code of the
new engine (probably under the GPL license). In principle I would prefer the
engine to be open-source, but the problem of clones makes the decision a bit
problematic.
In another recent thread, Joachim Rang liked the idea of open source, Marcus
Prewarski opposed it, and Tom Likens had mixed feelings about it. What do the
rest of you think?

I would like to see the source of a strong engine. And it is not
important what the code looks like. I know my own code that starts
nice and changes to #ifdef strange expression quickly. The things
I am looking for are new ideas and tricks for extensions, pruning
and other technics that are not published yet. Though I can and will
not implement everything in my own engine I will collect a stock
of ideas for my to do list.
But that may be bad for the development of own ideas. I do not know
whether this new knowledge will increase or decrease the probability
that I awake at night and write down a new algorithm.
I have seen the code of crafty and a few other chess programs, but I
am always afraid of looking too much. I don't want to copy code and
I don't want to create a clone of anything. I therefore prefer to
read ideas hints and pseudocode in papers and newsgroups. Then I can
try to adopt them in my program, make own experiments and decide
whether they are good for me or not. For example I have made a few
naive experiments with king safety, other evaluation terms and
pruning probability, without much success so far.
If anyone of the top programmers, that is everone with a better engine
than my Elephant, wants to read my source, that is ok. The price
is a few hints where the worst bugs are and what I have to do next.
Oh, that would delegate the hard work to others. And is it still my
own engine that is developed this way?
You see, there are not only problems with giving away source.
There are also problems with reading it.
Harald

[Tord Romstad]

Geschrieben von:/Posted by: Tord Romstad at 01 August 2004 09:45:11:
Als Antwort auf:/In reply to: Re: Open source or not? Opinions wanted geschrieben von:/posted by: Uri Blass at 31 July 2004 17:51:05:

Another possibility is to send the source code only to a small team of programmers that you decide to trust

This is also one of the possibilities I have considered, and it is very possible that I will
end up doing just this.
Tord

[Tord Romstad]

Geschrieben von:/Posted by: Tord Romstad at 01 August 2004 10:03:28:
Als Antwort auf:/In reply to: Re: Open source or not? Opinions wanted geschrieben von:/posted by: Sune Fischer at 31 July 2004 22:58:26:

I think that offering to open source the program is very generous, but I'm not sure if that's the best way to go if you're thinking about doing it 'for the good of science'.
Take Crafty for example, a very strong open source program, but kinda hard to read and figure out what's actually going on. Well, in my opinion anyway.
If I had my choice I'd rather read about the good stuff in a more compact and direct form, such as in an article. It's generally a bit easier than reverse engineering somebody elses code.
You deserve to become the author of a few popular and frequently quoted articles!
Changing the subject, what's that I read about you removing your attacktables?

I agree, but better yet is to have both. I think research papers are more interesting if the
program used in the experiments is available to the public, preferably with full source
code. It happens too often that some author writes about a new algorithm X, and
proves the effectivity of X by running tests suites and matches with a private chess
engine Y with and without the new algorithm. It is not unusual that nobody else is
able to reproduce the good results with algorithm X.
Would be fun, but unfortunately I don't have the time to write computer chess papers
at the moment. Writing a chess engine in itself is already time-consuming enough.
Just an experiment. I try to compute the attack information when I need it instead
of computing all attacks to all squares at every node. Attack tables sometimes give
me more information than I need, and sometimes not enough. It happens that I don't
need any attack information at all for some squares, and it happens that I only need
to know whether some square is attacked by (say) a white knight. On the other hand,
it also happens that I need more complex attack information than the attack tables
can give me. Therefore, I am now trying to compute exactly the information I need
at the moment I need it.
Whether this will be successful is too early to say. It is possible that I will give up and
return to attack tables.
Tord