+1 to Carsten's comment, specifically "Therefore the odd/even policy solves handling the differences between maven and OSGi versions and also keeps us away from surprises"
The odd/even versioning scheme was originally introduced due to Maven's special handling for the SNAPSHOT qualifier, which OSGi does not replicate. That leads to the situation where: - Maven: 1.3.0-SNAPSHOT < 1.3.0 - OSGi: 1.3.0.SNAPSHOT > 1.3.0 This has proven to be a nuisance during development, because once you install your new release, nothing gets updated and time is wasted finding out that the SNAPSHOT is still installed. IMHO there is nothing wrong with the odd/even versioning scheme and we should stick with it. We *could* switch it around to have odd release versions and even development versions, which would solve the eternal issue of "loosing" the micro "0" version (i.e. 1.3.0-SNAPSHOT (even) -> 1.3.1 (release). However, I believe that at this point this would cause more confusion than what it's worth. Regards Julian On Mon, Sep 23, 2019 at 3:32 PM Konrad Windszus <konra...@gmx.de> wrote: > > A workaround has been suggested by the bnd people and I have opened > https://issues.apache.org/jira/browse/SLING-8735 > <https://issues.apache.org/jira/browse/SLING-8735> to track implementation of > this workaround in the next bundle parent. > > Konrad > > > On 23. Sep 2019, at 11:41, Robert Munteanu <romb...@apache.org> wrote: > > > > On Fri, 2019-09-20 at 16:02 +0200, Radu Cotescu wrote: > >> +2 (as it seems that Konrad also agrees with this). Robert, will you > >> be the volunteer for proposing the change to bnd? > > > > I proposed the change > > > > https://github.com/bndtools/bnd/issues/3446 > > > > I know there is a lot more discussion, but this improvement will > > unblock us on the short term. On the longer run we could revisit > > odd/even versioning policy, but I guess that's quite a big change I > > would prefer to do it on our own terms, not forced by a tooling > > problem. > > > > Thanks, > > Robert > > >
