On 2008-01-15 at 13:44 -0800, Michael G Schwern wrote: > There are plenty of people out there which assume that Test::More, the > module which powers something like 80% of all Perl's testing, is unreliable > because it's version 0.74. The 0.x part says nothing about it's > reliability, it just means I'm too much of a wuss to just plant a flag and > say "this is 1.0" because of all the overloaded meanings of that version. > So that's why I say it can be actively damaging by giving the wrong > impression.
My stuff's all still 0.x simply because I can't be arsed to write the makefile/svn integration logic to handle >0 releases. For me, 0.x means that it's cut from at svn repository trunk at version x. 1.x.y means that I've switched to a formal maintained scheme because everything's stable. Hasn't happened yet cos I'm a wuss too. sieve-connect needs a little more interop testing yet. I also have software where the files come from a shared repository and so two hundred dot-releases might be entirely unrelated. But that's irrelevant since if x > y then x is newer and hopefully more bug-fixed than y. I preserve the ordering invariant. Meh, anyone come up with a decent set of integration rules for svn/make to be able to just tag a given repository version with the next release in whichever line, tracking versions along branches, etc? If there's something non-hateful pre-rolled, I'll go with that. -Phil
