Bug#144046: general: Sections are not finely grained
On Mon, Apr 22, 2002 at 04:15:45PM +0200, Javier Fernández-Sanguino Pe?a wrote:
> > I think it would be better to drop the sections altogether and use a
> > keyword-based system someone suggested a few months ago. The advantages
> > would be:
> > - ultimate fine-grainedness (?)
> > - no dillemas about where to put packages which fit in more than
> > section (like x11 net-related programs)
>
> Users need a hierachical layout in order to find software.
Who said that available keywords must be kept in a flat list? :-)
You could always group them in sections!
I think that a properly designed keyword system can have all the
benefits of a section-only system, while being much more flexible at
the same time.
> Keyword
> by themselves are not that much useful since they would be only appropiate
> to the language used.
What do you mean?
> Several disadvantages:
>
> 1.- more difficult to translate than sections
How is "modem" more difficult to translate then "comm"? :-P
> 2.- are not organised hierarchicaly (sp?)
Like I said, grouping them shouldn't be a problem. Then selecting a
group in a hypothetical package browser could mean the same as selecting
a "OR" of all the keywords in it.
> 3.- difficult to represent graphically in a package-administration gui
> (sections are easily represented as trees).
Ease of representation doesn't mean ease of use. The problem is with
packages which might be in more than one subtree (i.e. section): think
of a GNOME IRC client (could be both in x11/gnome or in net/irc in the
setup you suggest) or some network-based scientific computation system
(both in net and science/math).
Since a package may only be in one section, if you dive into wrong
subtree, you're moving away from the goal. This is the disadvantage of a
tree system.
But when using keywords, each keyword you select brings you closer to
the goal. You could think of keywords like of a tree system, where each
package can be at as many branches as many keywords it has.
I hope you know what I mean.
> If you want to have a keyword-based system I would suggest you
> take a look at dpkg-iasearch (yes, not documented, but it's a proof of
> concept) which uses natural language retrieval techniques (TFIDF
> vectorisation of documents) to allow for more powerful searchs than
> "apt-cache search XXX".
I'll have a look at it. But I think that such tools would be most useful
if each package would have a "Keywords:" field.
> Also take in account that the users will always see a
> hierarchicaly (sp?) division of software if using the menu system (in any
> window manager) or KDE and GNOME.
I guess that's a completly different thing. We're discussing package
management now.
regards
Marcin
--
Marcin Owsiany <porridge@debian.org> http://marcin.owsiany.pl/
GnuPG: 1024D/60F41216 FE67 DA2D 0ACA FC5E 3F75 D6F6 3A0D 8AA0 60F4 1216
--
To UNSUBSCRIBE, email to debian-devel-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Reply to: