Re: Future versions of teTeX, and TeXlive as a replacement
On Mon, May 29, 2006 at 18:47 -0300, Rogério Brito wrote:
>
> Sorry for the very late reply. I'm trying to catch up with d-t-m@l.d.o
> right now.
It probably is the right time to discuss these things again, now that it
is official that Thomas won't make another teTeX release.
> Yes, sure. The (latex) packages that I use the most made me pull in a good
> amount of (texlive) packages.
Let's get some numvers, first. Complete teTeX, that is -base, -bin,
-extra, and -doc but no -src is about 190M. The 'texlive' package from
versions 2005-2 draws in about 10 texlive packages that also sum up to
about 190M. However, the content of these packages is different.
> I think that something like a "popularity-contest"-like package for
> deciding what to include in the Debian packages would be a nice thing
> (and even recommend that to upstream).
>
> For instance, the packages that I use the most are (from memory):
>
>
> inputenc
> fontenc
> babel
> ifpdf
> geometry
> hyperref
> graphicx
> eso-pic
> amsthm
> amssymb
> amsmath
> xspace
> algorithm
> algorithmic
I think all but the last two packages are in tl-latex-base or
tl-latex-recommended, which is good, because all these packages are
important and often used. I think the seletion in these two packages is
allready quite good and based on popularity.
algorithm(ic) is more difficult, though, since it needs tl-latex-extra
which is larger than 100M. The other really large packages are
tl-fonts-extra (about 90M) and tl-doc-en (50M, installed by default!).
IMHO these three packages are candidates for further splitting and/or
reorganization. For example tl-doc-en is so large because it contains
lots of additonal documentation that might be interesting to some users,
but not as default (say tamethebeast, make-tex-work or examples from
various books). A splitting in tl-doc-en-base and tl-doc-en-extra might
make sense, but I haven't looked at the exact numbers here.
tl-latex-extra and tl-fonts-extra (aptitude has difficulties
scrolling through the really long descriptions) could be split up into
several packages. However, I don't think that popularity would be a good
meassure for these packages. I think some sort of 'amazon' approach (you
use this package, then you might be interested in that package, too;
sort of cross-correlation on popularity) would be more useful and in
line with the allready present splitting of other things into 'interest
groups'. However, this requires input from users. To use your example, I
have never used algorithm(ic) and have only a faint idea of what it
might be used for.
> I also like to use the following for changing things in my documents:
>
> mathptmx/mathpazo
> txfonts/pxfonts
> fancyhdr
All these are (or will be) part of the packages installed by 'texlive'.
I think that's fine, although i could live without [pt]xfonts, since
they are way to tightly kerned.
> euler
> beton
IIRC it would be better to use eulervm and ccfonts, both written by
Walter Schmidt who really knows how math fonts in LaTeX (should) work.
But those are scattered around, too.
This raises the question whether or not it would make sense to include
some sort of 'old, only for compatibility' section for things like
euler.sty. Would be useful IMO, but would also require a lot of work.
Anyway, the packages in Debian are closely related to the collections
provided by upstream. So any changes would probably be better done
there.
cheerio
ralf
Reply to: