Re: Automatic mirror detection
-----BEGIN PGP SIGNED MESSAGE-----
Miguel Ramos wrote:
> The solution you talk about (wich is software based) can be implemented at the
> network "layer". The solution is _anycast_.
> If we can manage to technicaly implement it, it would work as you propose.
I never implemented anycast, but my understanding of it is that - being
a routing scheme - it's not suited for connection oriented protocols
like TCP (and HTTP/FTP by extension). In practice this problem shouldn't
arise that often, but doing so (in a situation where the "nearest"
definition is dubious and the router hops constantly between two close
mirrors, for instance), it would be really annoying to debug/fix.
Also, the fact that not all out mirrors are interested or otherwise
capable of performing routing configuration changes on their networks
makes this a bit more difficult, since it would have to be AFAIK a
coordinated distributed BGP configuration, for instance.
Did I get your proposal wrong?
On a semi-unrelated note, I also just noticed there is a
pdns-backend-geo package, perhaps worth a look.
Cheers and thanks for the input!
Leo "costela" Antunes
[insert a witty retort here]
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
-----END PGP SIGNATURE-----