2 mirrorlist changes for this week

[Jeremy Katz]

On Mon, 2007-11-05 at 12:06 -0600, Mike McGrath wrote:
> Matt Domsch wrote:
> > On Mon, Nov 05, 2007 at 08:15:07AM -0800, Toshio Kuratomi wrote:
> >> Matt Domsch wrote:
> >>> 2) If a user isn't resolved by GeoIP, they will be given the global
> >>>   list, but without excluding mirrors who have an Exclusive Country
> >>>   set (e.g. Vasile's server really only wants traffic from .md users,
> >>>   but he sees global users).  This requires a change to
> >>>   mirrorlist_server.py trim_by_client_country() which I haven't made
> >>>   yet, but will look into doing so.  We should make this change to
> >>>   provide better accurate clients to our mirrors to keep their
> >>>   expenses down.  I'll post the patch here for review before implementing.
> >>>       
> >> I'm not sure how invasive this will be but it does sound like something 
> >> we want to keep our mirrors happy.  Maybe seeing the patch first will
> >> be better for deciding if it's too big a change during the freeze.
> >>     
> >
> > Here's the resulting function.  I think this is very clear code now,
> > and it works for me tested against live data on pt1.
> >
> > def trim_by_client_country(hostresults, clientCountry):
> >     results = []
> >     for hostid, hcurl in hostresults:
> >         if hostid not in host_country_allowed_cache or \
> >                 clientCountry in host_country_allowed_cache[hostid]:
> >             results.append((hostid, hcurl))
> >     return results
> >
> > Patch follows.
> >   
> 
> +1 from me, can I get atleast a +1 from someone else in sysadmin-main?  
> Its tested, low risk (as in we can roll back if there's any issues)

Looks straight-forward enough to me too.  And getting it in place today
does give us the time to know if something's wrong before Thursday.

So go go go! :-)

Jeremy