[pkg-bioc] Improving on our shared repository?

[Dirk Eddelbuettel]

On 20 June 2007 at 16:35, David Vernazobres wrote:
| > >I see. So we are having filesystem issues because we have too many 
| > >files.
| > >
| > >In that case I see no other choice but to split the 'one big dir' 
| > >approach into a tree of smaller dirs, maybe at the start 26 of them for 
| > >every first letter [ but there are too many with R, no ? ] so maybe we 
| > >need one dir per source package, or ...
| > 
| > 
| > why not use the filesystem structure used by debian for the package pools?
| > 
| > surely there's some code that can be borrowed for setting that up in a 
| > painless way....
| > 
| > --e
| 
| I was more thinking on the following: 
| * having 1 repository per archive (cran, bioc, omegahat). Ok then you
|   need 3 lines in your sources files, but you need to re-generate only
|   the repository where you upload (mainly divided the time by 2
|   (omegahat is really small)).
| 
| * and moving to a file debian archive structure. But as our packages all
|   start with r-cran-, r-bioc-,... I was thinking to moved to the 26 of 
|   them for every first letter after the dash (ie, 26 r-cran-[az] 
|   directories, 26 r-bioc-[az] directories).

The Debian structure is what I had in mind too -- but note how it has 
	-- 26 first letter
	-- plus lib* with another 26 subdirs
as there are so many lib* packages.

Do we have the same issue with r* ?   I am not sure we need to split for
r-cran-* as that name is the same as the source package, so let's just split
by first letter of source.

Anybody feel like doing a quick descriptive stat of how 'dispersed' those
first letters currently are?

| I was also looking at mini-dinstall for some time now, 
| as it's using a sqllite database and do not need a mysql or postgresql
| like dak. Or could we also have an postgresql access on Alioth (Dirk?)

Are you sure mini-dinstall uses that? I have been using mini-dinstall at home
for a long time, and set one up for local packages at work, and I don't think
I needed to do anything.  Which may of course prove your point of sqlite :)
I thought it just used cached file data...

And I am of course thrilled about Elijah offer to at least temporarily host
at IUB.  *If* we get up and running *and* manage to keep things up, I am sure
that other offers will follow.

Dirk

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