[Debtags-devel] Tags and facets

Benjamin Mesing bensmail@gmx.net
Mon, 07 Mar 2005 20:44:19 +0100


Hello,


> The implemented-in and langdevel tags are perfect in this respect
The uitoolkit facet is perfect too - I love it :-)

> So, this would shift the focus from understanding the packages, to
> understanding needs.  Which sounds much more promising, and has the
> potential for finding more useful categorization.
This make sense - and probably will obsolete the use facet..

> Other people, however, have wondered about other uses of tagging:
> tagging how a package is maintained ("orphaned-upstream,
> has-critical-bugs", ...), for example.  Facets in this case are
> essential to keep such other utility tags under a specific namespace.
> 
> 
> Given all these interpretations of tagging, the most interesting option
> is to leave it open for various groups to create facets and share them.
Please not! There is nothing worse than a individual vocabulary -
consider someone who makes a new facet. Of course he won't tag all the
17000 packages available so there will be perhaps 1000 packages tagged
with the tags from the new facet - leaving 16000 disregarded. So when
searching and using this tag, you will only get 1/17 of the results
(perhaps a little more because the percentage of packages carrying the
tag might be higher in the 1000 packages). Persons maintaining new
packages might not even have this facet available because they did not
added this source. This would lead to a lot of missing positives which I
my opinion is the worst error case that might occur.

Let other people make suggestions but finally all the tags should be
merged into a central database.

> The way tags are edited could change as well: tags in facets such as
> implemented-in and langdevel could be filled in by ordinary developers;
> tags in facets such as 'security' or 'field' could instead be filled by
> experts, who know the field and the software used in it.
I think all tagging should be left to the maintainers - they know their
packages best. Perhaps some "experts" could have an eye on some tags,
but don't make it a main responsibility.


> The other, is to design an interface which
> can present the user with a reduced number of facets.
Debian Package Search _can_ do that currently. You can use the Plugin
Settings to hide facets you are not interested in, and there is the
option to search through the facets. Are there other ideas how to do
this? I am willing to devote more time to implement this. Or perhaps
there is someone who wants to write another application? Mine is QT
based so there is the need for a gtk based one (every program must have
a QT and a GTK version :-)

> 
> These, and other like timestamping the tags, using RDF and semantic web,
> bayesian tagging over package metadata and documentation as well.
> 
> We are building quite some theory here, but it's mainly in our heads:
> when and how are we starting to put it down?
Come on, don't be unfair.. The debtags system is implemented and usable
by now. There is a UI for tagging and some for searching by tags. Even
the bayesian tagging was devoted some work! Don't talk us bad we are
good :-))

Btw. when will the debtags library report the short description for
facets? It looks stupid:
	Facet: admin Description: admin 


Greetings Ben