[tex-live] catalogue metadata usage in texdoc (was: texdoc index)

Reinhard Kotucha reinhard.kotucha at web.de
Sat Dec 31 03:09:45 CET 2011


On 2011-12-30 at 22:35:29 +0900, Norbert Preining wrote:

 > Hi Reinhard,
 > 
 > a lot of nice words and ideas like ...
 > 
 > > If texdoc is supposed to rely on the catalogue entirely, a few things
 > > have to be done.  First we need clear specifications.  Maybe we need a
 > > small working group.  
 > 
 > 
 > But .... can we take a practical approach? We do not need a tool
 > that in 100% gives the right thing, if we reach 95 or 99% that
 > would by far be enough.

Norbert, I tried to answer Joachim's question.  Hence, emphasis was on
the catalogue rather the current implementation of texdoc.

Furthermore, I suppose that if locales are to be supported, your
estimation of 95..99% will be too optimistic.
 
 > Now I don't have mycomputer with me, but I guess that if you write
 > a script that lists you for all TLpackages the file choosen by
 > texdoc, and make a statistic, I'm quite sure that there are a few
 > cases that are wrong. And foxing these cases by exception handling
 > is much easier than a *WORKING GROUP*!!!

Given someone writes such a script, wouldn't it be better then to use
the results in order to improve the catalogue data instead of just
adding new exceptions?  And don't forget Joachim's question:

 >> Why is the [catalogue] bonus just +1.5 and not more?


 > So, instead of habing written a long long email, why don't you
 > commit a fix for the malfunctioning of psnfss or vntex or both? It
 > costs a tenth of the time compared to your long email and fixes a
 > minimal annoyance instead of creating more and more complex things.

Seems that I wasn't clear enough.  I always want to reduce
complexity.  In my opinion, what we have now is too complex.  The
system depends entirely on heuristics and exceptions.

As far as vntex is concerned, there is nothing to do at the moment.
It works, but I expect problems if Manuel decides to support locales.
I can't do much before I know _how_ they are implemented.  And
regarding plnfss, the main problem is that if you fix a bug in
texdoc's scoring scheme, it has an impact on other packages too.
This is a big drawback of the current approach.

On the other hand, the catalogue provides pure data.  If it provides
all the data needed, texdoc doesn't need to rely on heuristics
anymore.  It becomes predictable.  Exceptions are then only necessary
if a user wants to change the default behavior.

Given that texdoc can rely on the catalogue entirely one day, it would
be simple and predictable.  It would be orthogonal too, i.e. if you
change properties of one package, it doesn't have an impact on otheres.
The scoring scheme has an influence on all packages.  If you fix a
bug, the whole world isn't as it's been before.

The reason I proposed a working group is simple.  We need a clear
concept before asking Robin to change or add anything.  There was a
discussion about documentation and localization a few years ago.  It
turned out that Heiko had much more experience and knowledge than
anybody else.  Thus, it would be silly to tell Robin what to do
without consulting Heiko before.

The expression "working group" might be interpreted as being a big
thing.  But what I meant is to discuss these things with people who
have more experience and knowledge than me before making a decision.
Simply in order to avoid having a specification which changes daily.

Norbert, I'm sure that you agree with me that it's fine that the
licence information in the catalogue is reliable.  Wouldn't it be fine
if other XML tags are reliable too?  

Let me summarize:  What we have now depends on heuristics and
exceptions.  It's not orthogonal, if you fix a bug here, you don't
know in advance what you break somewhere else. 

If texdoc could rely on the catalogue entirely, everything will be
easy and predictable.  *That* is simplicity, not what we have now.

Norbert, please don't blame me for making things more complex than
necessary.  You know me for more than six years now and you should
know that I've been preaching simplicity all the time.

Regards,
  Reinhard

-- 
----------------------------------------------------------------------------
Reinhard Kotucha                                      Phone: +49-511-3373112
Marschnerstr. 25
D-30167 Hannover                              mailto:reinhard.kotucha at web.de
----------------------------------------------------------------------------
Microsoft isn't the answer. Microsoft is the question, and the answer is NO.
----------------------------------------------------------------------------


More information about the tex-live mailing list