Subject: continual releases in computer chess
From: Vasik Rajlich
Message Number: 485975
Date: February 12, 2006 at 04:52:00
Hello,
my experience with various releases over the past two months has been quite positive, and one of the things that I have tossing over in my mind is the idea of continuing a sort of "rolling release schedule" for versions of Rybka after 1.2. Customers would pay for a subscription, rather than for a specific release.
Probably, the release schedule would be on the order of once per one or two months.
If this happens, of course existing customers will have their order turned into a subscription which starts when Rybka 1.2 is ready. In fact, I'd probably add some sort of a bonus for the early customers.
There are three problematic issues:
1) How to combine it in the future with CD sales.
2) How to convince someone whose subscription period just expired to renew immediately, when the next update will be only marginally better than the previous one. Basically, in that scenario, you are selling the future.
3) This whole approach depends on continual progress, and loses its point when progress comes to a crawl.
This last point is probably the most important. Note that it's not only a question of engine strength - for example, once the UCI-KNEX (knowledge extensions) are out, then a faster release schedule will again make more sense.
Naturally, this isn't strictly a computer chess issue. For a bit more eloquent general justification, please see here (scroll down to the section on "Releases"):
http://www.paulgraham.com/road.html
In the case of computer chess, here is why I think customers win:
1) The cutting-edge functionality is delivered immediately, rather than after a two to twelve month window of waiting for the next release
2) The final product is better, because
a) it is better-tested
b) it is possible for the developer to try more radical features and get feedback
3) It's more fun
There is also one additional nice point:
4) Piracy requires more work
Vas