Re: Versioning policy was Re: [PATCH] fix apr-config with symlinks

["William A. Rowe, Jr." <wr...@rowe-clan.net>]

At 07:42 PM 3/30/2003, Justin Erenkrantz wrote:
>--On Sunday, March 30, 2003 7:31 PM -0600 "William A. Rowe, Jr." <wr...@rowe-clan.net> wrote:
>
>>No... I like the every-other thought.  I'd go odds-devel/evens-release.
>
>And, what exactly is a odds-devel release?
>
>To clarify, what has been suggested for the odds/even policy is this:
>
>1.0.0: <initial>
>1.0.1: <devel>
>1.0.2: <release>
>1.0.3: <devel>
>1.0.4: <release>
>
>No statement on how compatibility is achieved in this model.

No, it's described below, but yes - the above is my ideal.

>The documented versioning policy we have says:
>
>1.0.0: <initial>
>1.0.1: <binary compatible with 1.0.0>
>1.0.2: <binary compatible with 1.0.0>
>1.1.0: <source compatible, but not binary compatible>
>2.0.0: <not compatible at all with prior releases>
>
>For a library, I like the last system by far.  -- justin 

It isn't contradictory, n.n.0 would be an initial drop, n.n.{odd} would be
the dev builds between qualified n.n.{even} releases.

Compatibility rules above would hold, and I don't even see a reason to
enforce httpd's 1.1 -> dev.  I don't see this library changing so much on
the main branch.  If I'm wrong, maybe we should consider this too, that
1.{odd}.* are all work-in-progress and don't follow the compat rules except
as applies to the prior 1.{even}.* releases.

Bill