On Sun, Jun 26, 2011 at 5:53 AM, Patrick Lauer <patr...@gentoo.org> wrote:
> So again, what are you trying to fix, and what makes you think it was
> broken to start with?

Well, I think there are things worth improving.  However, I'm not sure
that we should consider implementation of tagging a reason to
re-design the entire portage system.

I think the simplest way to go about tagging is to just stick tags in
metadata.xml and not change the way anything else works fundamentally.
 I could see a little room for allowing for future expansion, such as
defining namespaces in the xml.  Then you could have a set of
description tags, and later a set of file tags (a la portage file
search).

Lack of tools to do something with the tags isn't a problem - they'll
happen.  Probably the first place to do something with tags would be
packages.gentoo.org - if I want a quick list of all the text editors
on the system I don't mind doing a quick web search.

Sure, we can later expand things to provide a metadata index of some
kind and then command-line tools to search tags, or maybe a program
that sets up a tree full of symlinks or whatever after doing a sync.
I don't see why these have to be fully worked out to implement the
concept.

I think we should avoid changing the fundamental design of portage,
such as removing categories or allowing tags to be used as
dependencies/etc.  At least, not right now.  If we set up namespaces
for tags that might allow for such a thing in the future.

Also, there is no reason that tools couldn't help maintain the tags -
certainly if we used a tag namespace to list the files installed by a
package we'd want to automate that.  I don't think we should try to
tackle that out of the gates - for starters the list is
version-specific and not package-specific, so you'd either need to
figure out how to sanely build and maintain a union of everything, or
tag tags with package versions or something.

The main driver behind tags seems to be searchability, and I think
that is something we could easily improve.  I don't think we should
hold that up over an initiative to completely re-architect Gentoo...

Rich

Reply via email to