[texdoc] [tex-live] catalogue metadata usage in texdoc (was: texdoc index)
Manuel Pégourié-Gonnard
mpg at elzevir.fr
Tue Jan 3 22:31:54 CET 2012
Le 31/12/2011 18:46, Manuel Pégourié-Gonnard a écrit :
> The discussion about catalogue data and how texdoc is using it, should be using
> it now, and could/should use it in the future is very interesting, unfortunately
> I have to time to think and write an appropriately detailed reply right now,
> hopefully tomorrow or at most in a few days.
>
I re-read the discussion and here are my thoughts (summarized).
1. Regarding the fact that texdoc uses catalogue information via the TeX Live
database. I'm not aware of real-world cases where this is a problem. As always,
feedback is very welcome and if there are enough reports about wrong results
caused by outdated or otherwise wrong information in the texlive database, I may
switch to shipping information directly taken from the catalogue with texdoc.
As a matter of fact, in repackaged versions of texlive (such as found in Linux
distros), texdoc is already using metadata shipped with texdoc since there is no
tlpdb here. It should be no problem to generate this data drectly from the
catalogue and use it event in the "vanilla texlive" case. But I won't do it
unless I'm conviced by real-world cases that it needs to be done.
2. As Philipp pointed out, locales are supported since texdoc 0.80 (thanks to a
patch he provided, by the way). However, currently texdoc doesn't use language
information from the catalogue, it only guesses easy cases like finding
biblatex-de.pdf when looking for biblatex and the locale is de.
The reason for not currently using the catalogue here is again, the scoring
system is sort of a mess right now and I'm not quite sure how language should
interact with other criteria. Again, I plan to review the scoring system to make
it more flexible, and use language data from the catalogue at this point.
3. I don't agree with Reinhard's view of a purely catalogue-based texdoc for
several reasons. Granted, we all want simplicity and maintainability, but we
seem to diverge about what is the most simple and maintainable. I strongly
believe that the best approach, as Norbert said, is to have the script handle
something like 95% of the cases and fix the rest manually.
This is the approach currently taken with texdoc's alias file. One of my goals
is to improve texdoc so that we don't need too many aliases and the alis file
remains maintainable; for example, the first time texdoc started to use te
available metadata, it allowed me to cut the size of the alias file by 2 IIRC.
I'm not sure there is much to discuss here since this is also a matter of
personal preference. I know I won't change my mind on this point and Reinhard
probably won't change his either.
4. I've taken good note that Robin enjoys working on the catalogue and is open
to discussion if some more information in the catalogue is needed (which is
hardly surprising, but still good to hear). Currently I can't think of something
texdoc would need, but it may happen during the upcoming revision of the scoring
system, so I'll discuss this with the ctan team (and Norbert for the version in
the tlpdb) at this point if needed.
Manuel.
More information about the texdoc
mailing list