At 11:44 27/10/03 -0800, Iavor S. Diatchki wrote:
1. at what point should we remove the Unstable.* name, i.e. does anyone have any criteria besides "it seems to work"?
most software these days seems to be distributed on this basis, so perhaps we should adapt this as well. i don't have time to prove the corectness of the monad library, but it seems to work. i can't gurantee that there are no bugs, and i don't think we should freeze its interface as soon as the Unstable.* bit is removed (especially for some of the newer transformers). i think "releasing" the library is a good idea as it will give more people the opportunity to use it, hence find bugs in the implementation or the interface.
In the area of network protocol design, a touchstone often used is some variant of "two interoperable implementations" (e.g. the "running code" leg of "rough consensus and running code"). Maybe a comparable criterion for library code might relate to its use in real, complete applications? #g ------------ Graham Klyne For email: http://www.ninebynine.org/#Contact