Hi David,
On an unrelated tangent, let me say: darcs is cool :)
On Mon, 2003-04-28 at 13:51, David Roundy wrote:
> I would hope that rather than such generic terms, one could specify more
> specific tags for highly specialized packages and have these tags imply a
> certain degree of specialization. So this way a user interested in a
> specific specialization would be able to browse
>
> specialized
> bioinformatics
> physics
> mpb
> mpb-mpi
> chemistry
Well...there are already "chemistry" and "biology" tags. I do agree
that specialization::{high,low} or whatever isn't perfect, but having
both say "chemistry" and "specialized::chemistry" seems worse.
I guess it sort of depends too on how tags like "biology" will be used.
If say someone wrote a nice easy to use educational GNOME/KDE program
that provided an introduction to biology, it seems to me it should be
tagged "biology", but shouldn't be tagged specialized::biology.
What solution do you propose?
> Again, I would hope that a less general tagging of packages could be
> achieved, with the user level inferred from more specific tags. For
> example, it would be nice to have tags indicating the style of user
> interface a package supports. I imagine something like
>
> userlevel::novice = !specialized && (interface::gui || interface::curses)
Maybe. But your proposal above brings in almost all the packages.
Basically anything that's not biology or physics or whatever. That's
far from what I want. I mean, the difference for a novice user between
rhythmbox and (to pick a random example) mp3blaster is pretty large, in
my opinion.
This isn't to bash mp3blaster; it has a different (more technical)
audience.
Another example; Debian Desktop users, when searching for "spreadsheet",
should see gnumeric far before they see (e.g.) oleo. Now gui::gnome or
gui::kde might be a good first cut at an implication for
userlevel::novice, but both desktops have software which isn't targeted
at novices, so it isn't quite the same thing.
> where you'd probably want to include other tags that I haven't thought of.
> The only practical way I could see defining the userlevel of packages would
> be in terms of such a definition anyways, and this would make it easier to
> customize the definitions for meta-distributions (e.g. maybe you wouldn't
> consider any text interface to be novice-friendly--although this would make
> you leave out aptitude, which seems like a bad idea--on the other hand for
> a novice with noone handy to help them out, even aptitude would be
> confusing).
Right.
> I guess what I'm getting at is that I think that the tags should be as
> simple and obvious as possible, leaving more subtle distinctions as derived
> tags wherever possible.
I understand that goal. I hope my argument above is persuasive enough
to convince you that something like userlevel:: is a good idea.