On Friday 30 May 2008 22:28:37 Michael Croes wrote: > > There should also be some kind of hierarchy between tags, with the > > potential to be arbitrarily deep (even though most packages will probably > > use only one or two levels), so we can have, say, a dev/tool/scm/git tag > > for all git-related software, or games/roguelike/nethack for all NetHack > > derivatives. The package manager should probably be aware of the > > hierarchy, so a search for dev/tool/scm will find all dev/tool/scm/git > > automatically. > > I personally don't see the need for hierarchy when using tags. If you > want all tools related to git, search on the tags 'tool' and 'git'. > Having an hierarchy will in some way lead to the same issues a category > based approach gives.
It's not strictly needed, true, but it seems more organised to me. Possible benefits would be a less cluttered interface in some kind of graphical package browser (eg, only show the "games" tag at first, and expand the subcategories on request), and avoiding the need for developers to explicitly specify both the more general and more specific tags. On the other hand, it does make the tag names longer, which would go against that second one.... > > There would need to be some other way to > > disambiguate between packages with the same upstream name (but categories > > suck for that anyway, because it assumes that the two packages won't go > > in the same category, although it's plausible that there could be two > > dev-libs/libfoo or kde-misc/kbar). I'm not sure what that would be, > > exactly; the only thing that comes to mind is renaming one or both of the > > packages, but that's pretty nasty, so if someone thinks of something > > better I'd be all for it. > > The hard part about finding the right package is to tell the package > manager which one you've chosen. If it's just a package name that > identifies a package, then the identification has to be extended to > allow to differentiate between two equally named packages. Personally I > think in these days every package has it's own place on the net, so that > could be used as an unique identifier... Still sounds a bit sucky but > that's MY best idea. The thought did cross my mind... not /every/ package has its own page, though: sometimes upstreams have a single page for a group of related packages, sometimes we have multiple split packages for a single upstream package (I think we want to do that less in Exherbo than in Gentoo, but I doubt we can eliminate it completely), sometimes upstream websites die and the source is only available on Debian mirrors, etc, etc. There's also the possibility of sites moving around and being reorganised. _______________________________________________ Exherbo-dev mailing list [email protected] http://lists.exherbo.org/mailman/listinfo/exherbo-dev
