New Comps Groups

[Christopher Stone]

On 11/27/06, Bill Nottingham <notting at redhat.com> wrote:
> Christopher Stone (chris.stone at gmail.com) said:
> > If you ask what good it provides, then I have to ask what harm would
> > it cause?
>
> That's not how good engineering is done, generally - it's based
> on 'Why?', not 'Why not?' Any bit of new code:
>
> 1) can add bugs
> 2) adds a maintenance load
> 3) adds complexity
>
> The idea is to figure out the scenarios and personas you're trying
> to design for, and then figure out how to meet those needs.
>
> How does a checkbox list of 589 packages (roughly the number of
> perl-* packages) make someone's life easier? Is this better done
> via searching for 'perl' in the package search interface, for example?

I would say no.  Take for example python packages.  Some are called
python-* some are called py* some are called Py* some are called *py,
etc.


>
> Playing devil's advocate, if you go this way, wouldn't the better way be
> to automatically tag packages that install files in
> /usr/lib/python-2.4/site-packages as 'python modules', and generate the
> comps file from a database?

Yea, if you can autogenerate comps files that would be great!