[pkg-bioc] CVS %p

[Dirk Eddelbuettel]

On Fri, Apr 20, 2007 at 03:34:13PM +0200, Steffen Moeller wrote:
> Hi,
> 
> On Friday 20 April 2007 14:12:16 you wrote:
> > On Fri, Apr 20, 2007 at 12:22:33PM +0200, Johannes Ranke wrote:
> 
> > > Nice papar! I didn't know we arrived at 1700 packages autobuilt. What I
> >
> > Yes, David and Steffen did a *ton* of work this as astute readers of
> > the cvs list will know :)
> >
> > > don't understand however, is how the debian autobuilders for 11
> > > architectures will/would come into play. I think maintaining 1700
> > > packages in Debian would be difficult, even given the quality checks
> > > that e.g. CRAN R packages went through.
> >
> > We don't know either. I think the feeling is that that many packages
> > are not maintainable in the same way, so we probably end up offering
> > them somewhere else somehow. But first more/better automation...
> 
> Johannes has directly spotted one of the weaker points of our (accepted) 
> submission as it mixes two bits:
> 
> a) Debian is generally good for the provisioning of runtime environments since 
> the platform does not matter - the packages are available on them all
> b) We can seamlessly redo our automated builds on arbitrary platforms and 
> achieve the same cross-platformness even when not using the Debian 
> infrastructure directly.

Can we?  Or don't we have explicit/implicit dependencies of
"Debianness" in terms of 

  a) package availability, 
  
  b) naming (!!) [ almost every distro calls, say, Postgres libs
     something else ]
     
  c) infrastructure (pbuilder, ...)

> As Dirk pointed out, we are opting for b) with some bits in a).

Maybe we we are able to 'b-1)': redo on "arbitrary hardware provided
it runs Debian testing"  :)

As Henry Ford said: You can have the car in any color provided it is
black.

Dirk


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