Versioning

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

Golem

Re: Versioning

Post by Golem »

Ovyron wrote:
Carey wrote:You (and many others) dislike it simply becuase you expect the minor number to have only a single digit.
I think it looks better with one digit. But what about roman numerals? I don't think that they have been used before so it may be original.

I.i - I.ii - I.iii - I.iv - I.v - I.vii - I.viii - I.ix - I.x etc. (Zappa Mexico used the numeral II, for instance.)
Nothing new with roman numerals, the winboard chess engine Phalanx was using it for years (up to version XXIII if I remember correctly)... :D
User avatar
WinPooh
Posts: 267
Joined: Fri Mar 17, 2006 8:01 am
Location: Russia
Full name: Vladimir Medvedev

Re: Versioning

Post by WinPooh »

wgarvin wrote:Of course, what happens if you have 0.95 and you're getting kinda close to 1.00 but you don't think the program is mature enough yet to label it 1.00, because you want to save 1.00 for the first "non-beta" release? In fact, by 0.95 you might feel you are getting uncomfortably close and you might not want to go up to 0.98 or 0.99 because people will mistakenly feel that those versions are "almost, almost non-beta" even if that's not true?
I like Glaurung's numbering scheme. It was: 2 - epsilon/2, 2 - epsilon/3, .../5, ..., 2.0
User avatar
Ovyron
Posts: 4556
Joined: Tue Jul 03, 2007 4:30 am

Re: Versioning

Post by Ovyron »

Golem wrote:Nothing new with roman numerals, the winboard chess engine Phalanx was using it for years (up to version XXIII if I remember correctly)... :D
I stand corrected ;)
jswaff

Re: Versioning

Post by jswaff »

Bill Rogers wrote:Hi James
With the exception of items not related to playing strength I would say a major release would be anything that adds at least 100 elo points in playing strenght. That is just a person feeling though, I am sure there are a great many different ideas.
Bill
That's probably as good a system as any.

My engine (Prophet) is at version 2.0. Well, the latest public release is at 2.0 (about a year old). I went to 2.0 because I had literally rewritten most of the codebase. Some what "copied" directly from the 1.xx series, but it was "built from scratch."

The current codebase is already drastically different. It has a parallel search. So, regardless of how much improved it is in ELO, I think that merits a promotion to 3.0.

On the flipside of that, if I could somehow change a line of code or two to fix some bug and I got 100+ ELO out of it, I don't think I'd increment the major version number. I'd increment the minor version number.

So, I guess I'm in the camp that uses the version numbers to track changes in the code, not the playing strength.

--
James