New Comps Groups

[Toshio Kuratomi]

On Thu, 2006-11-30 at 21:21 -0500, Jeremy Katz wrote:
> On Thu, 2006-11-30 at 11:32 -0800, Toshio Kuratomi wrote:
> > I'm writing an email application and want to find a library to help me
> > do signing and encryption.  Do you know if Fedora includes one?
> >
> > I'm looking for something that I can use to make rpms in a clean chroot.
> > Is there a package for something like that in Fedora?
> > 
> > I'm working on some surveying software and I need to do a lot of work
> > with maps, coordinate plotting, etc.  What libraries are available in
> > Fedora to help me get my work done?
> > 
> > There is software to answer all these questions in Fedora but it's not
> > listed in comps.  Just look at the repoview page and you'll see why
> > Ungrouped leads to a horrible end-user experience.
> 
> For the first and third, though, you don't want to browse.  Because if
> the granularity of the groupings (which you browse) is good enough that
> you can say "oh, that's obviously the only group I need to look in" you
> either have a) 100s of groups or b) 100s of packages in each group.  And
> that's never good for finding things.  Instead, you want good search.
> 
The question that Jesse asked was "why is NoGroup bad"?  My response was
the Nogroup URL for repoview and a list of things I'd like to find in
them.  Without allowing categorization of all packages good browse
becomes impossible.

Browsing can be superior to search.  Many times you don't quite know
what you're looking for.  You have a few key thoughts about it and by
browsing you can add other criteria that help narrow the focus to what
you truly want.

Your granularity argument isn't thought all the way out.  If you have to
solve the problem of browsing packages you can come up with some ideas
that don't have the dichotomy of "too many groups vs too many packages
per group":

* Use hierarchical groups.  Have a managable number of toplevel groups
which each branch to a manageable number of subgroups.

* Think thesaurus.  There are millions of words in English.  To find
exactly the one you want, you start with a word that means something
like the one you are looking for.  Then you read that section, find some
other words that add some of the nuances you want and read the sections
for those words until you find the word(s) that convey the exact feeling
you desire.  For packages, you have an entry point of a rough category
(cryptography) which you'd want to browse for further categories (email,
signatures, library, C) which you continue to browse until you have a
small enough selection of packages that you can review them and find one
that fits your needs.

> For finding mock, browsing _could_ get you there, but if you know that
> specifically what you want, again, searching is always going to get you
> there faster.
> 
I was looking for plague, not mock.  If there were good groups, then I
could have found plague and mock close to each other and realized that
plague was what would fit my needs.  If you were the search engine's
algorithm, this is a case where browse would have been better than
search because it would allow me to find things related to the keywords
I was using.

-Toshio
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://listman.redhat.com/archives/fedora-extras-list/attachments/20061130/ab95940d/attachment.sig>