I think one good real-world test would be to actually run the frozen code on twistedmatrix.com for a day before releasing. This would of course only be a supplement to other harder tests, but as this incident shows, it would be a useful one.
On Tue, May 08, 2001 at 06:52:41PM -0500, Glyph Lefkowitz wrote:
This whole incident indicates the need for a better testing procedure; now that people actually rely on twistedmatrix.com to do stuff, I think that the release process should take a few days in order to ensure that *all* aspects of the code are tested. Proposals for something more formal will be gladly accepted :)