[Nice] Gracefull fallback, renegotiations?

[Olivier Crête]

On Fri, 2008-07-04 at 09:29 +0300, Rémi Denis-Courmont wrote:
> On Thursday 03 July 2008 19:25:13 ext mikhail.zabaluev at nokia.com, you wrote:
> > >Will there be always only one socket per component/transport/af? So
> > >whichever local candidate we pick doesn't matter since they'll all be
> > >sending the exact same packets?
> >
> > Don't you have to bind to every local interface, in order to send candidate
> > checks properly? It might matter with VPN and IP migration.
> 
> Well, we _could_ sendmsg with IP_PKTINFO/IPV6_PKTINFO ancilliary data to set 
> the source address per-packet. Also on Linux you can re-bind UDP sockets.
> 
> So we _could_ deal with multiple network interfaces with only one socket per 
> component/transport/family. Considering that we only have one transport 
> (UDP), and that (on Linux) you can send IPv4 datagrams on IPv6 socket, we 
> could in principle do with a single socket per component.
> 
> Kai and I had been thinking it would be better to support multiple sockets 
> because ICE-TCP would definitely require it, anyhow. But but... ICE-TCP does 
> not seem to be going anywhere at the moment, due to poor support from the 
> NATs. And we don't have a "customer" for TCP at the moment (file 
> transfers??). And ICE-TCP is way more complicated than ICE-UDP (that tells 
> something!), so I guess we'll skip that for the time being.

I guess even if we have a TCP socket, nice could guess which local
socket to pick based on the forced candidate...

-- 
Olivier Crête
olivier.crete at collabora.co.uk
Collabora Ltd
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
Url : http://lists.freedesktop.org/archives/nice/attachments/20080704/18ede010/attachment.pgp