Bojan Smojver wrote: > On Mon, 2009-12-21 at 17:01 -0600, William A. Rowe Jr. wrote: > >> I don't think Bojan actually suggests we nuke it. >> > > No, no nuking at all. Just bit of skipping. > > >> The suggestion is that all n.even releases would be ABI stable. n.odd >> releases would be purely alpha- and beta-, with clear notices to the >> user of what they are obtaining, and that updating is likely to break >> their applications which are ./configure'd --with-apr-dev or some >> similar flag. >> > > Yep, something like that. In other words, if we have say 1.8 released > and we are working towards 1.10, then anything in 1.9 is completely > API/ABI unstable and relying on it being present in 1.10 is a gamble. Of > course, the relationship between 1.8 and 1.10 would still obey the same > rules as we now have for say 1.3 to 1.4. >
-1 ... I've always found that approach a bit weird. There's no functional difference between 1.(oddx).y and 1.x.y-dev. It makes marginal sense to use the odd/even approach in projects like the Linux kernel where one expects to have dozens of intermediate development "releases" that do need to be differentiated. APR is quite a bit smaller and there's less flux, so changing the versioning scheme would cause unnecessary pain to users who're already well accustomed to it. Besides, I'm aware of quite a few projects who use the same versioning scheme and refer explicitly to our versioning.html file; if we change that without making a huge announcement, we intercourse those projects too. So I agree with Rüdiger that we can't actually change what APR version numbers mean before 2.0. If we /do/ happen to make a 1.x.y-dev pre-release that doesn't fly and we're concerned about breaking early adopters, and want to really differentiate between different releases, we're allowed to do a 1.x.y+1-dev without a 1.x.y release in between. This is going to be rare once the -dev and non-dev libraries become mutually unlinkable, so I really wouldn't worry about burning too many version numbers. They're free, after all. -- Brane
