[issue9025] Non-uniformity in randrange for large arguments.

[Antoine Pitrou]

Antoine Pitrou <pitrou at free.fr> added the comment:

This wouldn't be the first time reproduceability is dropped, since reading from the docs:

“As an example of subclassing, the random module provides the WichmannHill class that implements an alternative generator in pure Python. The class provides a backward compatible way to reproduce results from earlier versions of Python, which used the Wichmann-Hill algorithm as the core generator.”

Also:

> FWIW, we spent ten years maintaining the ability to reproduce
> sequences.  It has become an implicit promise.

IMO it should either be documented explicitly, or be taken less dearly. There's not much value in an "implicit promise" that's only known by a select few.

(besides, as Terry said, I think most people are more concerned by the quality of the random distribution than by the reproduceability of sequences)

----------
nosy: +pitrou

_______________________________________
Python tracker <report at bugs.python.org>
<http://bugs.python.org/issue9025>
_______________________________________