On Thu, Jun 26, 2008 at 03:28:21PM -0700, Dan Price wrote: > On Thu 26 Jun 2008 at 05:23PM, Nicolas Williams wrote: > > On Thu, Jun 26, 2008 at 03:12:53PM -0700, Dan Price wrote: > > > On Thu 26 Jun 2008 at 05:01PM, Nicolas Williams wrote: > > > > One good reason to separate pkg and software version numbers is that you > > > > might make changes to a package without changing the packaged software > > > > at all. Then, if your intention had been to have the pkg version number > > > > mimic the packaged software's, how would you encode such a change into > > > > the package's version number?? > > > > > > You'd republish the package with a newer timestamp. > > > > Like SVR4 packaging's PSTAMP? I don't think that's a good idea. But > > Can you cite a single reason why this mechanism, which is working > perfectly fine, and which we're already using, is not acceptable?
Yes: a) it's not very user visible; b) it's not meaningful to anyone looking at it (PSTAMP is at least much less meaningful that a version number, and quite possible meaningless without reference to a history of published packages with the same version but different PSTAMP); c) it can't be used to express dependencies; and d) I don't like it :) > > quite apart from that, if pkg version numbers be constrained to integer > > tuples then I think we'll be better off having pkg version numbers bear > > no resemblance to the packaged software's version numbers. I strongly > > recommend that; alternatively, make pkg version numbers strings (but I > > agree, implicit numeric ordering has desirable properties). > > Feel free to make a concrete proposal by getting involved and working on > the code. As it stands now, you're describing a project other than this > one. I'm proposing no code changes, thus I have none to contribute. I'm proposing a change in use. _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
