On Fri, 2004-04-30 at 14:53, Pekka Savola wrote:
> On Fri, 30 Apr 2004, JORDI PALET MARTINEZ wrote:
> > The point is the case when I travel to an IETF, for example, and
> > I've native IPv6, but the performance is very poor. I can always opt
> > for using a better transition mechanism (example a "near" TB/TS).
>
> FWIW, I'm concerned about methods to "optimize" poor native IPv6
> connectivity. We need to fix the native connectivity instead... :-/
Well there is this issue of 'policy' which the IETF can't fix.
For instance most of the 'educational and research' networks/NRENs
currently have IPv6 deployed and working but if you have the following
situation:
Commercial ISP <-> { IPv6 Internet } <-> NREN
Your traffic will most likely flow like:
Commercial ISP -> Transit -> NREN
but the return route will be:
NREN -> 6net -> Commercial ISP
And guess which route the "Transit" networks take. Good example was the
"Global IPv6 Launch Event" in Brussels. Connected to Belnet, thus all
the NREN connectivity is good. But if you wanted to reach a host in the
Netherlands all the traffic would be routed directly over 6net ->
Surfnet to the Commercial ISP but all the return traffic isn't accepted
by 6net thus that was routed 'nicely' over the US...
Nice policy, but as mentioned nothing the IETF can fix except for what I
did: clickety click in my heartbeat tool and tada a perfectly working
tunnel to a dutch Commercial ISP as IPv4 didn't have the
IPv6-over-the-US problem.
This might be solved one day however when IPv6 gets more deployed and
the routes get shorted, but this example for instance makes it quite bad
when interacting between Commercial networks and NREN's. Notez bien that
between Commercial networks or between the NREN's most connectivity is
quite good actually.
> On the other hand, there _are_ a few practical ways for optimization
> which are rather simple: a "local" 6to4 relay, and a "local" Teredo
> relay. Those can significantly help when you're on native v6 but
> communicating with someone who isn't.
Doesn't help when the policy is stupid ;)
Greets,
Jeroen
Attachment:
signature.asc
Description: This is a digitally signed message part