On Thu, Jan 20, 2005 at 10:30:03AM +0100, Andreas Barth wrote: > Hi Joel, > > However, I think this feature still makes sense: > > For example, autobuilders need an unique distribution name to decide > where to upload packages to. So, e.g. for dak.ganneff.de, uploads may > also go to unstable-ganneff; they are however mapped to unstable. From > the users point of view, unstable-ganneff doesn't exist. The situation > for volatile is similar: Uploads to stable end up in stable-volatile - > but, there is for obvious reasons no symlink there.
One additional question - do these distribution entries need Release
files, and if so, what should be placed in them? Normally a Release
file for sid will have the name 'unstable' in it, for example; since
these are not supposed to be user-visible, and Release files always are,
should the Release file not exist at all, or should it be filled out
with the target distribution name?
Hmmm. That leads to another interesting question, as well:
Can a "virtual" dist (say, unstable-ganneff) target a symbolic name
("unstable"), or must it target a codename that appears as a primary entry
on the filesystem ("sid")?
My initial thought on this would be...
"Virtual dists" must always target a codename that exists in the "dists"
hash. No symlink is created, and they are not considered when building
Release files (which will determine what name is used entirely from
'dists', as they currently do).
Thus, if you have the following dists:
stable => woody
testing => sarge
unstable => sid
experimental => experimental
Then you could do any of the following in virtual_dists:
unstable-ganneff => sid
my-broken-packages => experimental
However, if you wanted to point testing-volatile at sarge-volatile, you
would first need to add sarge-volatile to the dists hash; it couldn't exist
*only* in virtual_dists.
Does this make sense?
--
Joel Aelwyn <[EMAIL PROTECTED]> ,''`.
: :' :
`. `'
`-
signature.asc
Description: Digital signature
