[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: mirrorlist cgi



On Mon, 2006-07-17 at 20:27 -0500, Josh Boyer wrote:
> On Mon, 2006-07-17 at 21:08 -0400, seth vidal wrote:
> > On Tue, 2006-07-18 at 01:59 +0100, Andy Burns wrote:
> > > seth vidal wrote:
> > > 
> > > > That's completely correct.
> > > > 
> > > > there are no UK-specific mirrors in the rawhide mirror list that are
> > > > current. so it won't hand any back to you.
> > > 
> > > oh fair enough, but how does it "know" if 
> > > http://www.mirrorservice.org/sites/fedora.redhat.com/development/x86_64/
> > > is up-to-date or not?
> > 
> > 
> > Read the check-mirrors.py script in the cvs link I pointed to earlier.
> > 
> > In short - if it doesn't have the same repomd.xml as the mirror master
> > does then it's not current.
> 
> Forgive my complete lack of mirror knowledge, but are mirrors told to
> rsync this file after the rest of the repo is updated?  Otherwise I
> would think you'd still catch mirrors in the middle of a repo update if
> not.

nope. A mirror could be in the middle of a sync. We cannot control HOW
the mirrors sync the data. We can't know. But this keeps users from
being bounced to sites with multiple, out of date metadata.
Additionally, if they're given a whole set of mirrors it means that the
mirrors they get will be 'roughly' more current then never pruning the
mirrors at all.

The only other option we have is checking every file and that will take
WAAAAAY too much time and bandwidth.

I'm open to other suggestions.
-sv



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]