On Wed, Aug 8, 2012 at 10:13 AM, Toshio Kuratomi <a.badger@xxxxxxxxx> wrote: > On Wed, Aug 08, 2012 at 09:22:44AM -0400, Andy Grimm wrote: >> Hi, all. I have a package up for review, and there's some question >> about how it should be versioned. The package is stax-utils ( >> https://bugzilla.redhat.com/show_bug.cgi?id=794923 ), which has the >> following situation: >> >> * They have released in the past using date-based versions (20070216, >> for example) >> * The POM files which I was able to find (this is a java package but >> does not bundle its own POM) use the date-based version. >> * I'm pulling the latest version from the subversion repository, >> because I need something newer than the last stable release. >> * There's nothing in the source tree to indicate a version at all >> >> The pieces of data that I have to use in a version-release string are >> the date: 20110309 and the svn revision: 238. I think the correct >> thing here would be to call it 20110309-0.1.svn238 , which indicates >> that it's a "pre-release" of 20110309, even though this specific >> version is unlikely to ever be released. Does this make sense? >> >> Part of what's in debate here is that the package guidelines indicate >> that the date should be in the release tag, but this seems redundant, >> and there appears to be precedent (apache-commons-csv, batik, vpnc, >> etc.) for omitting it. >> > Upstreams change the way that they are versioned from time to time. For > that reason, I would use: > > Version: 0 > Release: 0.1.20110309svn238 > > This protects you in case something happens with upstream and they start > making releases with more traditional MAJOR.MINOR format. > > (This sort of thing can happen if someone new takes over upstream, for > instance.) Fair point. This doesn't really jive with the maven versioning of the package, but since nobody from the Java SIG has weighed in with an alternative suggestion, I'll take your advice. At the very least, it's easier to go from 0 now to the date later than it would be to go the other way. Thanks. Andy -- packaging mailing list packaging@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/packaging
