package metadata

Janusz S. Bień jsbien@mimuw.edu.pl
16 Apr 2003 06:41:49 +0200


On Wed, 2 Apr 2003, Enrico Zini wrote on debian-devel:

>The initial main areas of interest were:
>
> - status of package metadata: what is ok, what is still missing, how could it
>   be added;
> - package searching and browsing techniques, to cope with the large and
>   increasing number of packages;

No long ago I have made two exercises: finding all packages related to
Japanese languages and all packages related to TeX (the second one
would be easier if I was aware of
http://debian.vitavonni.de/packagebrowser/). 

I will describe them on some other occasions, now I would like to
share with you the following observations:

`kakasi' package is ,,kana to kanji simple inverter' and contains in
particular 5 programs with UNDOCUMENTED dummy man pages
        /usr/share/man/man1/kakasi.1.gz 
        /usr/share/man/man1/mkkanwa.1.gz
        /usr/share/man/man1/atoc_conv.1.gz 
        /usr/share/man/man1/rdic_conv.1.gz
        /usr/share/man/man1/wx2_conv.1.gz 
While the dummy man page for kakasi direct correctly to
/usr/share/doc/kakasi/ for more information, the page for
e.g. mkkkanwa is misleading, because it refers to the non-existing
/usr/share/doc/mkkanwa/ directory.

It seems highly desirable to differentiate the UNDOCUMENTED man pages
by including in them at least the information about the package the
program belongs to.

Let's now consider another example:

`chasen' package contains a Japanese morphological analyser,
documented in Japanese

`lookup-el' package is a general purpose dictionary lookup tool,
documented only in Japanese.

I think that the package metadata should distinguished clearly these
two cases, using e.g. the standard language codes: the language of the
application domain and the language of the documentation.

Best regards

Janusz

-- 
                     ,   
dr hab. Janusz S. Bien, prof. UW
Prof. Janusz S. Bien, Warsaw Uniwersity
http://www.orient.uw.edu.pl/~jsbien/