Re: Policy for packages providing newer CTAN versions than teTeX
Hi, Frank and other maintainers.
I'm sorry for the late reply, but I had some personal problems (including
being submitted to a surgery). I am now slowly catching up.
On Jun 25 2005, Frank Küster wrote:
> Rogerio Theodoro de Brito <rbrito@ime.usp.br> wrote:
> > I think that I can prepare a Debian package of algorithms. We would
> > only need to see what dependencies and conflicts would be needed (or,
> > perhaps, diversions/alternatives---I don't know).
>
> We have not yet decided how this should generally be done. We plan to
> support an additional TEXMF tree where packages can install their newer
> versions and shadow the ones from teTeX. You could install into this
> tree, and need not care about teTeX versions.
Ok, that's one possibility. But I would also like to use one package that
is only provided in teTeX 3, namely, xkeyval.
For the sake of "backwards compatibility", I am only using keyval and it
works fine so far, but I think that xkeyval would be better.
> However, there is also a different approach. While sarge was prepared a
> couple of LaTeX things that weren't in teTeX-2.0 were packaged for Debian
> and are provided as separate packages. teTeX-3.0 now contains these
> packages, and to avoid duplicates we simply don't install them from
> teTeX. We could, in principle, do the same with algorithms.
Yes, I see. Like lmodern, which I heard/read is available in teTeX 3.0, but
which wasn't replaced by the version of teTeX in experimental (I'm using
that now --- the only thing I miss regarding teTeX 2.0 was the
debconf-driven way of choosing language hyphenation).
> As I said, we have not yet discussed whether we want to adopt a general
> policy here, and which. I could imagine the following approach:
>
> - Usually, packages that provide files not in the current teTeX install
> themselves into /usr/share/texmf
Ok, that's a more or less sensible way to do it, as teTeX as a whole has a
lot more inertia to change and a smaller package may be updated more
frequently than the whole teTeX distribution.
> - Usually, packages that provide newer files with older versions already
> in teTeX install into /usr/share/site-texmf/
>
> - If a new upstream version comes out, we "delete" files from teTeX in
> two cases:
>
> a) If the package is new in teTeX and already packaged for Debian
> (into /usr/share/texmf)
>
> b) If the package already was in teTeX, but now has an established,
> well-maintained version living in site-texmf.
I don't know exactly what you mean by "deleting" (if you really mean
deleting the files), but this would be a nightmare for the
tetex-maintainers if more packages actually got packaged for Debian.
Perhaps, just a way of overriding what is already installed (say, via an
alternatives/diversions method) would be a saner way of dealing with the
case.
This way, if the user don't install the newer package from Debian, having
the one from teTeX being used seems to be a good compromise.
I am sorry if this has been discussed to death and if you already have
decided on a solution to this or if there is any obvious failure in what I
said.
> Unfortunately, site-texmf is not supported in unstable (and even not in
> the current experimental version. So if you get packages ready soon,
> you (or rather your sponsor) can upload them simply with files in
> /usr/share/texmf, but please notify us. If we are faster with uploading
> a teTeX-3.0 with site-texmf support to unstable, better put them there.
Well, due to the problems I already mentioned, I didn't have the
opportunity to package it for Debian, only to release a newer version on
CTAN.
Thanks, Rogério Brito.
--
Rogério Brito : rbrito@ime.usp.br : http://www.ime.usp.br/~rbrito
Homepage of the algorithms package : http://algorithms.berlios.de
Homepage on freshmeat: http://freshmeat.net/projects/algorithms/
Reply to: