[pkg-bioc] Re: Genetics Program

[Dirk Eddelbuettel]

On 13 December 2005 at 13:52, Steffen Moeller wrote:
| > And even that is still very murky if the files come from a .debian.org
| > server. People will get confused.  So I think this needs a little more
| > thought and a clearer definition of what standards if any we are shooting
| > for.  What is you goal?  Pass 'R CMD check' (a given) but also pass
| > lintian?
| Lintian should not complain about very hazardous conflicts with Debian policy. 
| The built packages however should reflect what you, Dirk, sketched as a 
| Debian policy for R.

That document is getting pretty close to being irrelevant as it is two years
old. Rewriting / updating it may be a good first step towards specifying what
we want to accomplish here though.

| > What you mean "for sponsoring" ?  I cannot really sponsor any more packages
[...]
| This is fine since for an alioth repository there is no sponsor required. I 

Ack.

| Different wording: The BioConductor folks have shown that an automated 
| compilation (with the getBioC.R script) can produce something useful and we 
| should strive for something useful in a completely automated fashion for 
| systems running Debian, too. If we cannot get this automated building 

That is where we may differ. I think it is much harder than what BioC does,
but also potentially much more powerful.

The BioC people have it much easier -- eg RdbiPgSQL "just assumes" Postgres.
We know we have it, but we need to express which version(s), potentially
which configure flags etc pp. Likewise, our package names are not their
packages names.  They can say "Depends: affy" but affy isn't a Debian or even
Alioth/Debian/BioC packages (which would be r-bioc-affy).

So we all agree: thanks to the CRAN build infrastructure of R packages,
reused at BioC, and the coherent layout and content of their archive, we can
attempt autobuilds ... but it requires further work and infrastructure at our
end.  I would like some clearance on what infrastructure we're going to put
in place.

Given that we are talking about hundreds of packages, I think it pays to
think in terms of infrastructure.  Now, I could agree that we may just want
to start on a subset of 20-40 BioC packages to get things rolling, get some
sort-of-automated builds for i386 and amd64 etc.  But we need to keep the
bigger picture in mind too.

Dirk, off to work

-- 
Hell, there are no rules here - we're trying to accomplish something. 
                                                  -- Thomas A. Edison