Re: listing IPv6 support on mirror pages

[Mattias Wadenstein <maswan@acc.umu.se>]

On Fri, 27 Jul 2007, Simon Paillard wrote:

> On Fri, Jul 27, 2007 at 09:58:22AM +0200, Mattias Wadenstein wrote:
> > On Thu, 26 Jul 2007, Bas Zoetekouw wrote:
> >
> > > There is a request in the bts [1] to show which mirrors are reachable
> > > over IPv6.  I've already added a new field "IPv6" to the mirror
> > > masterlist [2], but Simon Paillard suggested on irc that that might not
> > > be the best solution, because there could, in principle, also be
> > > ipv6-only mirrors, and because people might be relying on the existing
> > > format of the mirrorlist.
> > >
> > > Does anyone have a suggestion on how to best solve this?  Is it ok to
> > > add extra field to the master mirrorlist and the mirror lists on the
> > > website?
> >
> > This seems to be a display issue only, if it has IPv6 it has an AAAA
> > record, if it has IPv4 it has an A record. Just do a resolve at
> > website generation time?
>
> The choose-mirror d-i component relies on the "source" file
> Mirrors.masterlist to build the list of mirrors displayed at install.
>
> So if we want do discriminate ipv6 capable mirrors at the install phase,
> we need to put the details in the source file itself, not only in one of
> the compiled html file.

So use dns and Mirros.masterlist for choose-mirror compling too?

> Moreoever, it may happen (it already happened actually) that there is a
> AAAA record but no working service behing it (while the A record is OK).

Then the mirror is broken and needs to be fixed. Just like a mirror that 
only serves 404 pages instead of debs.

> So I don't think testing only the dns is enough reliable.

Putting this data on yet another place makes it much less reliable since 
it is more likely to be outdated in at least one of the places.

I'm strongly against a manually edited field in Mirrors.masterlist when 
the authorative source of the data already is in DNS.

/Mattias Wadenstein