[pkg-bioc] CVS %p

[Steffen Moeller]

Heyho, some slight disagreement :-)

On Sunday 29 April 2007 17:54:36 David Vernazobres wrote:
> On Sat, Apr 28, 2007 at 08:36:59PM +0000, CVS User smoe-guest wrote :
> >  * could we somehow automate an upload of the erected packages to a
> >    shared repository on alioth? Or should we have a cron job on alioth
> >    that scans our build directories every 6h for novel builds and pulls
> >    them via wget to update the packages list afterwards?
>
> hum, well this will mean adding few Giga on the alioth hard disk space.
> So let's say, that ~1G per arch (ok there are some binary independent
> package, I know). So about ~3G for i386/AMD64/PPC....
> We may need to add some more check : only upload the clean litian|linda
> packages. I am not sure that the way of going (today), need more
> description and discussion...
I think differently here - why does it need to be perfect? We are not even 
uploading to experimental and we are at least as good as anybody installing 
the packages with traditional means intrinsic to R.

The DFSG-hurdle as imposed by lintian and co should be the technical hurdle 
for uploading to unstable  ... not that we'd ever get to unstable in the 
first place for all packages ... and we certainly attempt to reach the 
compliance for all packages.  However, in my view we should neither hide 
imperfect code nor imperfect packages. There is no adequate way to 
communicate back to the community than with the packages we produce - hence - 
they have to be somewhere no matter really in what quality with respect to 
the standards of Debian.

Your main concern so I presume is with the many GBs we'd occupy. My 
understanding is that GBs are cheap, only the persistant backup costs dearly. 
I would be prepared to fund a 250GB external disk if anybody wants that. .... 
75€ at ALDI or PLUS or SATURN or so these days.

As a start I would be happy with a directory that is not backuped. 

> >  * shouldn't we add this shared repository to our pbuilder sources list?
> As soon as we have one and that it's working, yes !

Let us come up with a toy system for a few packages and then see how it 
scales. My current vision is that all of us place our packages symbolically 
linked to a web server and wget retrieves everything newer than what is 
available on the server, updates the PACKAGES.gz afterwards. This way we can 
also upload new packages any time manually without interfering with anything.

The number of architectures I see at the moment is two: amd64 and i386. This 
is ok to me even if we do not share packages for the first few days until we 
come up with some extra intelligence.

Many greetings

Steffen

-- 

Dr. Steffen Möller
University of Lübeck
Institute for Neuro- and Bioinformatics
Ratzeburger Allee 160
23538 Lübeck
Germany
T: +49 451 500 5504
F: +49 451 500 5502
moeller at inb.uni-luebeck.de