[Pkg-octave-devel] octave-forge

[Rafael Laboissiere]

* Thomas Weber <thomas.weber.mail at gmail.com> [2007-09-22 15:30]:

> Am Sonntag, den 16.09.2007, 14:38 +0200 schrieb Ólafur Jens Sigurðsson:
> > > But again: which problem is the local compile supposed to solve?
> > 
> > Well, I was trying to solve the problem of not having a debian package
> > for the forge, a bit megalomanic of me :-)
> 
> The idea of having octave-forge as individual packages and pkg.m was the
> easy installation of packages by a local user. So, what good are
> distribution packages for if you can easily do it by from inside Octave?
> Like the extensions for Firefox, or whatever.

There are many arguments supporting the generation of individual Debian
packages for each pkg in octave-forge:

1) Site-wide installation: Debian packages put the files in a place
   accessible to all users of the system.

2) Dependencies: libraries will be automatically installed as needed.

3) Integration with Debian: instead of relying on pkg.m, we rely on dpkg/apt
   for creating/removing files from the system.  That sounds the obvious
   thing to do in Debian.
   
> > Just out of curiosity, what are the problems that you see in packaging
> > the forge?
> 
> Try doing it for the 40+ packages and wait for the first change that
> inflicts all packages (such as a new Octave version).

Well, the Debian CRAN/BioConductor Group [1] is doing fine as regards
packaging the individual packages of R [2].  Dirk Eddelbuettel have given a
talk [3] ar the useR conference [4].

[1] http://wiki.debian.org/?AliothPkgBioc
[2] apt-cache search r-cran | wc -l
[3] http://user2007.org/program/presentations/eddelbuettel-1.pdf
[4] http://user2007.org/program/

I still have the hope that we will sort out the problems and find a way to
(semi-)automatically generate Debian packages from octave-forge components.
However, I have not plunged deeply into the problem, so I trust Thomas'
judgment about the difficulties.

-- 
Rafael