On Sun, 19 Dec 2004 03:44:49 +0100 (CET), Dag Wieers wrote: > On Sat, 18 Dec 2004, Jeff Spaleta wrote: > > > On Sat, 18 Dec 2004 22:55:26 +0100 (CET), Dag Wieers <dag@xxxxxxxxxx> wrote: > > > Jeff, give an example where it confuses the version comparison or shut up. > > > > Shall I construct an example using rpm -Fvh using packages using the > > zork and zelda disttag thrown into a directory? the distrotags do > > affect comparison if the distrotag continues to be a part of the > > release tag. > > This is of no value as I explained before. Remove both zork and zelda from > the release-tag and there still is no good reason to prefer release '3' > over release '2' since there's no relation. > > Similarly how would you decide if zelda or zork should be used. There's no > logic to it. That's why the repotag is at the end, if it's up to the > repotag to decide what to happen it's already a lost case anyway. > > Thus the release tag has little value if you have different repositories > without relation or coordination. And even with some coordination it may > not matter (as Michael pointed out). Unfortunately you twist my words here. Referring to release (!) tags, I asked why release 3 from repo A should upgrade release 3 from repo B? It only does because a repo tag is included in the release tag and becomes the most significant portion when the rest is equal. When that happens (and not only then) we have a problem. You know my view on repositories which upgrade eachother of overlap eachother in an undefined way -- and I don't really wish to pound on it endlessly. > Repotags are only involved when mixing repositories and in those cases the > release tag has limited use. When repositories are mixed and the repo tag of multiple versions of a package becomes the least significant part of RPM version comparison, than that's the least important problem (but still an issue which is beyond the scope of this thread).