Re: [gentoo-dev] Re: Two-level portage tree is irrelevant.

[Richard Freeman <rich0@gentoo.org>]

Christian Faulhammer wrote:
> Hi,
> 
> Dmitry Grigoriev <dimgel@mail.ru>:
>> Jeremy Olexa already answered me in bugzilla that this is not new
>> idea, but I'll submit my suggestion here anyway as a "voice of
>> crowd". :) I'm just home user with about 2 years linux experience, do
>> like gentoo, but with exception of this inconvenience.
> 
>  People already suggested tagging which embraces your proposal and is
> more flexible.  The problem is: Somehow has to do that.
> 

I have a few thoughts on this (which are a bit scattered as they cover a 
few different related topics):

1.  I'm not a big fan of extracting metadata from file names or paths in 
general (no, I don't want to rehash EAPI-in-filename here).  I think 
that files should be named and organized in a way that makes sense, and 
that all metadata read by the package manager should be stored IN the 
file, not in the name or path.

2.  Each package should have a unique identifier that doesn't change 
(much).  Currently package category/name satisfies this.  It could do so 
better if we separated the naming from the classification (tagging/etc) 
since then we wouldn't have so much renaming going on.

3.  There should be other ways of finding packages that are more 
flexible - no restrictions on unique names, or being in a single 
category/etc.  Tagging sounds great for this.

4.  There should also be ways of storing ownership/support of packages. 
  Again, these should be flexible, and the current xml approach works 
pretty well, but tagging could also be used for this (maybe with 
additional fields).

In terms of how to maintain all the tagging, I can think of a few 
approaches:

1.  Introduce the feature and just let devs have at it, and see what 
happens.  Initially it won't be any worse than what we have now, and 
maybe it will become useful.  This won't really cost anybody that much time.

2.  Have some dev take the lead on it who is interested.  Sure, it takes 
time that could be spent on other things, but I don't see volunteer 
effort as being a zero-sum game.  For all we know the dev who takes this 
on was otherwise going to quit out of boredom.

3.  Find ways to let non-devs contribute.  This could either be in a 
centralized or non-centralized fashion.  You could even have a tagging 
system where people add tags via some webpage and vote on tags others 
have added.  If some team's ebuilds get tagged as "buggy" they can 
ignore it or use it to identify areas for improvement as they see fit.

You are of course right that some devs at least need to build the 
infrastructure to make all of this work (whether in portage, or some 
outside application, or whatever).  That doesn't mean that devs need to 
do all the reorganization work...