Re: Sharing Secrets (was: [computer-go] Computer Go hardware)

["Frank de Groot" <frank@xxxxxxxxxxxxx>]

Sorry guys, I should have been more specific.

What I meant by "secret" was:

- not previously published (eg. David has never made a secret of anything 
IIRC)
- useful for the majority of Go programmers (eg. the general architecture of 
a massive multi-CPU system is not really very useful for the average Go 
programmer)
- It has to be a "competitive secret" (as I thought was clear from my post).

Meaning, it's a secret in the sense of: "I will never tell because people 
will DEFINITELY 100% SURE get a benefit out of this if they would adopt it.
Meaning it is a mathematically proven certainty, not a design philosophy, 
however good it works out in practice.

An example:

I say I use 64-bit Zobrist hashing.
OK, that's already some info, like almost everyone knows what Z-hashing is 
but some might not know that 64 bits is really needed for Go 
positions/larger patterns.
Still I feel comfortable to share because if someone would implement 64-bit 
Z-hashing it would would very badly without taking care of a certain aspect. 
That "certain aspect" is virtually unknown to all Go programmers except very 
few. I wouold be willing to wage a bet that no Go programmer has taken care 
of that very important aspect and I think 64-bit hashing simply doesn't work 
in a scaled Go program without it. Meaning, it's an A-superplus secret..

Now, the sulution to this "aspect" is found very, very sparsely in 
literature (but you might have to search for weeks even with targeted 
searches) and the best solution mentioned I could find is still sub-optimal 
and only experimentation and more literature research and talks with 
colleagues finally yielded the "perfect" solution that I think could 
contribute greatly to everybody's Go program.

That's what I mean by "secret".
I have made the mistake once to tell an important (and useful for him) 
"secret" to a Go programmer whom's help I wanted and he simply said "Thank 
you" and I got nothing in return. So I will not throw my little inventions 
over the internet but as I said, when people are willing to share I will do 
that too because the synergy outweighs the disadvantage (for the 
commercially inclined).

But that means I want to see a few people coming with really key elements 
that can be proven to be mathematically sound, are totally unknown for 99% 
of us, almost impossible or totally impossible to find in literature 
(meaning, they are original, patentable inventions!) and are clear to all to 
be very useful. But ideas are a dime a dozen, I'm looking for exact, 
working, implementation descriptions of highly powerful, proven techniques 
that are simple to implement and (almost) completely new or at least very 
unknown in comp. Go. 
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/