Re: [gentoo-dev] Do I really need the tree?

[Wazow]

Chris Gianelloni <wolf31o2@g.o> writes:
> Anyway, for now it is much simpler to have a tree of ebuilds which are
> easily maintainable than a single (or a few) large xml files which would
> become a maintenance nightmare for all the developers involved.
> Currently there are many developers who work on only one ebuild in a
> particular area.  As a good example, I maintain exactly one ebuild in
> app-emulation.  What kind of separation would there be for the xml
> files?

Chris, I think you have confused things a bit. I believe that the OP
complained about the amount of useless information being
kept/transferred via portage tree, while only package names, versions and
names/versions of dependencies are needed to compute the set of
installed ebuilds. I guess nobody questions ebuilds. Only it seems way
to precautious to keep all of them on the local drive, instead of just
downloading those you need, once you call emerge package-name.

I am not a great fan of xml, but whatever format it would be it could be
automatically generated from the portage tree perhaps several times a
day (and then propagated to mirrors). Mirrors would still probably need
to keep the hole portage tree, but rsync would be performed only on the
packages you need at the point of emerging them. There seems to be some
restructuring in the code, but most of the mechanisms we have now would
need to be kept. Dependency calculation would need to use the new format
though and the batch generating the xml-thingy would need to be written.
Also not much (if anything) would change from developers
perspective. You would just work with ebuilds as you do today. Local
portage overlays could still be supported.

I actually like this idea a lot.

Andrzej


--
gentoo-dev@g.o mailing list