how the mirror submissions are handled [was Re: problem with Debian mirror submission mechanism (fixed now)]
On Mon, Jan 29, 2007 at 03:12:49PM +0100, Bernhard 'elven' Stoeckner wrote:
> > If any other masochists exist out there who would like to
> > join me in this tedious little job, please feel free to let me know.
> *holds up a hand*
> Glad to hear the form works again. :)
> How may I help?
The parsing of mails consists of two main things - parsing mirror
submissions, and parsing random queries.
Hopefully the bulk of the job would be mirror submissions. Previously, they
used to come only to the mirrors alias, so those would have to be extracted
from the mailboxes; newer ones come into the bug tracking system under the
'mirrors' pseudo-package, so they are available to all.
The parsing of automated submissions is generally straightforward - you
paste the new entry at the end of Mirrors.masterlist (which is located in
webwml CVS, in the directory webwml/english/mirror/) and run make full >
full.html, and see if it 'compiles' - if it's parsable. Then you verify the
existence of links, most simply by clicking through from a browser. For
rsync URIs you have to copy&paste, or at least I don't seem to have seen a
better URI handler from a browser.
You can often run into broken formatting, and it's not uncommon to have
second-guess the submitter, by looking around their site. After you get to
the root of the tree, you have to check project/trace to see if the mirror
leaves a proper trace. If not - mail the submitter/admin. Then it's
necessary to take a peek into a randomly selected pool subdirectory in order
to find out which architectures the mirror carries - for example, I often go
to pool/main/g/gentoo/ (should confuse whoever is reading the logs ;).
Based on that, you add an Archive-architecture: field in the stanza for the
If the addition to Mirrors.masterlist is successful (where success = basic
functionality), you commit it to CVS and move on. If not, remove it and
don't commit it.
So, anyway, if you're still interested :) let's just wait a couple of days
to see if anyone else volunteers. Then we'll divide the load properly so
that there is no duplication of work, and start gnawing the backlog.