> Antoine, there does need to be repeatablity; there's no question about
> that.

Well, that doesn't address my proposal of making it repeatable accross
bugfix releases only. There doesn't seem to be a strong use case for
perpetual repeatability.

If some people really need perpetual repeatability, I don't understand
how they can rely on Python anyway, since we don't make any such promise
explicitly (or do they somehow manage to read in your mind?). So,
realistically, they should already be using their custom-written
deterministic generators.