Re: Licsensing scientific data

[Sylvestre Ledru <sylvestre@debian.org>]

Hello Andreas,

Le dimanche 31 juillet 2011 à 10:04 +0200, Andreas Tille a écrit :
> - Sylvestre, did you by chance saved the gobby
> file because gobby.debian.net was switched off 

I guess you saw that since then:
http://debblog.philkern.de/2011/08/debconf11-gobby-documents.html

As attachement, you will find the one relative to the debian science
roundtable.
I will try to update the wiki soon.

Cheers,
Sylvestre

* Interaction of pkg-scicomp
 - NMU upload + svn migration to complete it
 
* Advertising on Debian Science
 - how ?
   * ?blog: e.g. Debian-Med got wonderful one http://debianmed.blogspot.com/ , NeuroDebian got some too ;-) : http://neuro.debian.net/blog/
   * ?presence on scientfic conferences (e.g. NeuroDebian does that on neuroscience-related conferences)
 - New nice website
 - to who ?
 
* List software. Update blends tasks:
 - missing software in the archive
 - available packages in the archive but not listed in blends
 
* BibTeX files
 - please consider adopting/extending http://git.debian.org/?p=pkg-exppsy/debian-bibliography.git
 - unify/work-together with http://wiki.debian.org/UpstreamMetadata
 - Debian references: continue M.Hanke-initiated http://wiki.debian.org/CategoryPublication : TODO - wiki2bib fetcher/validator
 - proposals:
   * debian/copyright (bad idea)
   * debian/control (bad idea with screams)
   * debian/bib (in BibTeX format)
   * debian/upstream-metadata.yaml (1-1 mapping with bib)
   * use RDF (or have some converter from .yaml)

* More fine grained tasks
 - Ideally aim at selection being done by sub-tasks within d-i
 - warning: we could lose critical mass

* Giving credit to upstream

* Sloppy backports?

* Build-time testing

* Helping upstream
  - expertise transfer among projects
  - build-time testing (together with automatic sloppy backports provides lots of feedback)

* Enable pinning to defined versions of programs
 - usually people want to work with a specific version of scientific program;
   how can we address this issue? Should we?
 - this is quite a new concept in Debian, because some of these users would
   really like forwardports of old software (to ensure reproducibility)
 - but if reproducibility isn't preserved in newer version, there must be something
   broken somewhere!
 - snapshot.debian.org can be used; this mean that old branches are not updated
   anymore
 - however, having multiple branches in Debian requires LOTS of work
 - static versions of packages are not a good idea. (Security) bugs will stay
   and the user might not be aware. 
 - users can also set up virtual machines to have full reproducibility of
   calculations (having the same libraries could be insufficient: you have to
   have the same whole environment)

* DebTags
 - create tasks from debtags
 - we need more fine-grained tasks to match the two classifications

* Usage of the mailing list
 - Proposal by Sylvestre Ledru to separate users and packaging questions
 - Argument by Adam Powell IV that "normal" users will ask on upstream MLs or other mailing lists