Re: Intent to package: molecular biology programs
On 28-Aug-1998, Stephane Bortzmeyer <bortzmeyer@pasteur.fr> wrote:
>
> [Important note: unlike many such announcements here, this one is really
> an intent. I've done almost nothing yet.]
>
> I'm thinking about packaging for Debian several molecular biology
> programs. AFAIK, there is only one, among the hundreds of programs in
> daily use by the biologist, which is debianized (Rasmol, by John Lapeyre
> <lapeyre@physics.Arizona.EDU>). We work here with *many* different
> programs, which are typically painful to compile, install, upgrade,
> remove, etc. So the Debian system could help.
>
> The prime candidates would be Phylip and Clustal (also Blast, may be).
>
> The problems I see:
>
> - most of these programs are awful, from an Unix point of view. No man
> pages, no command-line interface ("For such analysis, type 1, for such,
> type 2"), horrible coding. It is not seriously possible to clean them,
> we'll have to live with it. Is it a reason to exile them in contrib?
Only if they cannot be clean up (e.g. licenses are useless and depend
on non-free stuff).
>
> - most of these programs have retentive licences, often not because the
> author was opposed to free software, but because he thought he was able
> to write a licence and the resulting text is both undecipherable and
> retentive.
>
Common problem.
Best to contact the author, and suggest a re-worded license that is
more clear. Also let them know that as it is it cannot be put into
the main Debian system, thousands of users, and so forth.
> - many programs, since they come from different origins, have the same
> name. At least three "scan" executables to put in /usr/bin, besides MH's
> scan. Should I create a /usr/gensoft tree, like we do here to prevent
> name clashes? (Yes, I know only X11 does that on Debian, but the set of
> all biology programs is almost as large.)
If they are not user executables, they can be put in a
/usr/lib/packagename directory.
If they *are* user executables, they need to be separated. A common
technique for doing this is to create a shell script named after the
package, which executes the correct command (which lives in
/usr/lib/packagename)
For example
cvs add
cvs admin
cvs annotate
cvs commit
etc
So you could have
phylip scan
phylip create-genetic-mutant
phylip fuse-genome-with-fly-genome
phylip create-replicant
etc..
> - some of these programs are mostly used with *very* huge databases of
> genomes, which are unlikely to be found on any PC (Debian or else)
> machine. Anyone here works in biology?
Well, if they database is really huge and unlikely to be available to
anyone but the people running it, releasing the Debian package isn't
much use (although creating it might be helpful for local use).
Unfortunately, biology is falling into the trap of making genomic
information proprietary (and incredibly, genomic information taken
from nature may be made proprietary) so it might be difficult for
anyone to use such programs simply because there are no suitable
data-sets.
> May be I could draw experience from the people who package other
> scientific sets, but there seems to be very few domain-specific programs
> in Debian?
At the moment that is true, but there are people interested in improving
this. I think licensing has been a bit of a problem so far.
--
Those who would give up essential liberty to purchase a little temporary
safety deserve neither liberty nor safety. - Benjamin Franklin
Tyson Dowd <tyson@tyse.net> http://tyse.net
Reply to: