Re: Misclassification of packages; "libs" and "doc" sections
Daniel Burrows wrote:
>
> Could you elaborate on this? I know the current situation is confusing,
> are you trying to say something more? (or are you just agreeing with me?)
>
I just agree with you. I'm a long time debian user, but still have
a lot of difficulty browsing the archive.
> True. Emacs is an editor, a mail client, a news client,
> a web browser, a Lisp interpreter..argh..
>
An entire computing system carefully disguised as a text editor.
> Sorry, you've lost me. Could you define TUCOWS, please? (it doesn't
> have anything to do with bovines, does it?)
www.tucows.com -- a software repository that windows users tend to use
a lot. Similar in nature is winfiles.com
Well is has something to do with bovines, just look at the site :)
>
> Also, are you disagreeing with me, or just restating what I said more
> succinctly? It looks like the latter..
>
I agree with you about the need for a hiearchy. I might have said it too
succintly :)
Well, I disagree with you in that the hierarchy should be richer
than a tree. In an is-a hierarchy with single inheritance, the whole
hierarchy is a tree (in the formal sense), and an is-a hiearchy
with multiple inheritance is a directed acyclic graph.
X11_Client Development_Tool
\ /
\ /
Graphical_Debugger
/ \
/ \
[DDD] [xxgdb]
You get the idea...
> > A part-of hierarchy is also useful. In this sense, the system may
> > be partitioned into subsystems. Base may be one of these subsystems,
> > X windows another. If a suitable part-of hierarchy can be attained
> > this can be used to distribute release management as well (in that
> > subsystem decomposition is modular)
>
> I'm not sure I see why this would be useful except for the sake of
> finding out what subsystem Debian thinks a particular package goes in.
> This is usually the last thing I care about when looking for software.
>
An example is release management. If subsystem boundaries are small
compared to inter-subsystem dependencies then a subsystem can be
released as a sub-distribution; the release management can be decoupled
from that of other sub-systems. Also each subsystem may have different
formal requirements, in other words this is for implementation of
the distribution, not the interface :) A user would be less
interested in this.
> Hm, could you provide references for the research? I think some
> cognitive science stuff could be useful as well; I should probably spend
> some time next weekend reading up on it.
>
A friend gave me these URLs to have a look at, but there are
certainly countless resources on the net. I'll take a
look at some ontology-based classification systems, too.
http://ksi.cpsc.ucalgary.ca/KAW/
http://www.smi.Stanford.edu/people/musen/index.html
Here are some of the papers I found interesting
http://citeseer.nj.nec.com/55740.html
http://citeseer.nj.nec.com/farquhar96ontolingua.html
http://citeseer.nj.nec.com/33789.html
http://citeseer.nj.nec.com/labrou96semantics.html
http://citeseer.nj.nec.com/luke97ontologybased.html
http://citeseer.nj.nec.com/artale96partwhole.html
http://citeseer.nj.nec.com/guarino94ontology.html
These are all relevant so don't omit any of them, some of these
are particularly elucidating for the task at hand. At least
read their abstracts. :) If you'd like, these can form
a frame of reference for our discussions. [I spent a lot
of time to find these .. :) ]
I approached the subject from AI perspective, of course
you may examine the way other disciplines have approached
the same problem. Although AI is considered to be a part
of cognitive science, other researchers may have differing
points of view. CS approach is usually more likely to
be implementable though :) [ There are times when you
read an entire book on linguistics and still not be able
to write a single program that implements the theory.
Especially if the author's from the Chomskyan school ;> ]
Thanks,
--
Eray (exa) Ozkural
Comp. Sci. Dept., Bilkent University, Ankara
e-mail: erayo@cs.bilkent.edu.tr
www: http://www.cs.bilkent.edu.tr/~erayo
Reply to: