On 07 Mar 2015, at 2:40 PM, Jeff Trawick <[email protected]> wrote: > Pull this small amount of code into httpd, perhaps as a private interface for > 1-2 modules that need it. Let it improve with fewer constraints. Future APR > 2.0 will improve from the relatively unfettered changes, and httpd 2.4 users > won't have speedbumps introduced by dependence on an evolving skiplist > implementation. > > (Keep the skiplist code in APR trunk up to date with changes needed for > httpd. The APR stable releases can pick up compatible code fixes, but that > probably won't be a priority for anyone but non-httpd consumers, and I'm not > aware of any such people working on skiplist thus far.) > > Thoughts?
We need to address the fact that a non compliant skiplist implementation exists in v1.x of APR. Skiplists are formally defined, and we are not doing APR users any favours by providing an API to end users that we know is broken. I think we must fix what we have in v1.x, following our versioning rules, and deprecate any broken behaviour. While I agree that making skiplist stable in httpd would have been helpful, we didn’t, and we can’t go back and change that now. Regards, Graham —
