Re: IPv6 tunnel over NAT

[Keith Moore <moore@cs.utk.edu>]

> > well, obviously it depends on the applications being run and the
> > quality of the underlying links.  if you're running real time apps
> > over UDP over PPP over TCP over a lossy IP link, you'd probably
> > be much happier using Teredo.  do we really want to say, for instance,
> > that it's okay for the generic solution to break streaming audio?
> 
> a) Begs the question of whether any of our transition tools are robust
>    enough to support serious real-time streaming use.

6to4 has a fixed overhead/delay per packet.  Teredo and NATs probably do also.
 
> b) PPP/TCP looks like a reliable link with very erratic timing
>    characteristics.  I'm fairly certain that unidirectional streaming
>    protocols can deal with that via buffering (so that it's not real
>    time anymore).  Whether the deployed implementations can or not is
>    another matter.

PPP/TCP considerably amplifies the amount of jitter in the underlying
network.  In general, apps that are jitter sensitive will do worse
with PPP/TCP (assuming any significant jitter or packet loss in the
underlying network) than any per-packet tunneling or translation mechanism.

Keith