Re: Splitting med-bio: real soon now.
On Sun, Aug 08, 2010 at 08:43:30AM +0900, Charles Plessy wrote:
> I wonder if we need some transition, like having the old med-bio package
> depending on the two new ones, before being removed one release later?
That's an interesting question. Generating appropriate metapackages is
quite simple but I care about a sensible solution for the tasks package.
in the science-biology task file there is the following code:
Depends: med-bio
Meta-Depends: svn://svn.debian.org/blends/projects/med/trunk/debian-med/tasks/bio
While the Meta-Depends is not (yet) interpreted, but the idea behind
this was to "resolve" the dependencies to list all the packages
mentioned in the referenced task file. So you are able to list the
whole content of another task on the web page - which makes IMHO more
sense than just listing the metapackage itself. If we decide to go
with the med-bio wrapper, I should really start coding to implement
this (3 years old) plan.
> For the names of the new packages, we can either clearly differentiate them, like:
>
> - med-bioinformatics: Command lines tools for bioinformatics
> - med-graphical-sequence-analysis: Graphical tools for sequence analysis and molecular biology
Is graphical sequence analysis no Bioinformatics?
> or we can have one graphical and one command line package with the same base name.
>
> - med-bio-headless: Command line tools for bioinformatics
Headless?? I have no association with the term headless. I'd prefer console.
> - med-bio-gui: Graphical tools for sequence analysis and molecular biology
>
> A third possibility, perhaps closer to Steffen's proposition, is to have the gui
> package include the headless one:
>
> - med-bio-headless: Command line tools for bioinformatics
> - med-bio: Tools for bioinformatics and molecular biology
>
> (in all the cases, since the scope of the med-bio package grew beyond
> microbiology, I recommend to take the opportunity to change its description).
While I perfectly agree that we should split med-bio I'm not really
comfortable with the split according to the user interface. IMHO this
is a bit orthogonal to our tasks approach: Tasks should contain
packages which are useful to work on certain problems. They do not
differentiate on how to *use* these programs (console or gui). IMHO
we have even prepared a better way for a split:
$ grep -i "X-.*category" projects/med/trunk/debian-med/tasks/bio | sed 's/^X-.*Category: *//' | sort | uniq
Analysis of RNA sequences.
Annotation
Assembling
Clustering; Gene expression data
Clustering; SAGE expression
Comparative genomics
Gene prediction (through GHMM)
Genome assembling
Genome Browser
Genomics; Clustering analysis (+GUI)
Genomics; Codon usage analysis
Genomics; Prophage detection in prokaryotes
Graphical representation of sequence conservation
Molecular modelling and molecular dynamics
Motif detection
Multiple genome alignment
Phylogenetic analysis
Secondary structure of nucleic acids
Sequence alignments and related programs.
Sequence assembly
Sequencing
Target duplex prediction
Tools for the molecular biologist.
tRNA discovery
Visualisation
So I would suggest to rather split somehow according to the content of
the package. These categories do not all make real sense and should be
stripped down to about four or five and as I explained in my summary of
the Debian Science BOF at DebConf[1]. For instance in my institute it
would make sense to have some overview about "Next generation
sequencing".
Kind regards
Andreas.
[1] http://wiki.debian.org/DebianScience/ProblemsToWorkOn#Morefinegrainedtasks
(not finished at the time of writing, but this topic is covered)
--
http://fam-tille.de
Reply to: