Re: debian package browser

On Thu, Jan 29, 2004 at 06:28:09PM +0100, Erich Schubert wrote:

 Hi,

> You could also install synaptic-debtags. IIRC that does build all
> subgroups, not hiding small ones like my does.

I tried, and that's "better" (though not "perfect"), indeed.

> > So the goal is to have approximately the same number of packages in
> > subgroups as the number of "flat" packages displayed.

> No, the goal is to keep the "flat" number below a treshhold like 20
> packages. As long as this is not fulfilled, a subgroup is formed by the
> following rule:
> For all matching packges not yet in a subgroup, their tags are counted.
> the tag with most packages, but not above ~80% of them (to avoid
> unneccessary depth in the tree) forms a new subgroup, removing matching
> packages from the "remaining" set.

> This "look only at packages not yet in a subgroup" will avoid nested
> subgroups, such as "ip networking" included in "network and
> communication" - iff an "network and communication" subgroup is formed,
> there won't be an "ip networking" subgroup.

And why not?
One could want to search "network and communication", then "ip
networking"...
Yet, searching "ip networking" would imply "network and
communication".

> > I think the result is not always the best strategy. Suppose you know
> > exactly what kind of functionnality you're looking for (and that's
> > most of the interest of this interface)... then you're often forced to
> > wander through the "flat" list of packages, and sort out many packages
> > which don't correspond to what you're looking for.

> It's often easier to check a list of 10 packages than checking 5
> subgroups for 2 packages each.

That's not my point of view.

> > I think that my search would more efficient with a constant strategy.
> > Anyway, one can provide both in the future, like I said.

> As long as you know exactly what you need and that there is a tag for
> it.

Yes. And for me it's the interest of the thing.

> The package browser aims for people not knowing that.

?
And what are they supposed to be looking for, then?

> Others could just have entered IDS into apt-cache search.

Oh no. I was to say that tags are so much better that "apt-cache search",
because the string you are looking for is not obligatorily present in
the description.
I was to say this, when I tried "apt-cache search ids", to see what it
brings. And the result is self-explanatory. Just try it.
For me, it gives 115 packages, 95% of them having absolutely nothing
to do with detection intrusion. And it doesn't even give snort, but
snort-doc! It could even have not given it at all... and that's
logical.
Tagging is far more superior to a silly string matching.

When I'm looking for an IDS, I want to choose several tags, which
finally give me the one I want : "Intrusion Detection System", and get
presented a list of _all_ the packages that deal with intrusion detection,
and maybe some more "subcategories" which enable me to refine my search a
little more (dameon, library, etc,....)

That is clearly the best way of doing things, in my opinion.

 Hervé

-- 
 _
(°=  Hervé Eychenne
//)  Homepage:          http://www.eychenne.org/
v_/_ WallFire project:  http://www.wallfire.org/