option 212 for 6RD

[Ivan Pepelnjak]

BR reassembly is a patently bad idea from the cost/performance perspective. 

OTOH, DS-Lite does handle the MTU issues by performing B4/AFTR fragmentation/reassembly, assuming that the upstream packets are usually small. Good luck with BitTorrent ;))

Ivan

> -----Original Message-----
> From: Mikael Abrahamsson [mailto:swmike at swm.pp.se]
> Sent: Saturday, January 19, 2013 11:47 AM
> To: Erik Kline
> Cc: Tore Anderson; Ivan Pepelnjak; Daniel Roesen; IPv6 Ops list
> Subject: Re: option 212 for 6RD
> 
> On Sat, 19 Jan 2013, Erik Kline wrote:
> 
> > But if the tunnel pretends the MTU is 1500 and fragments such a packet
> > into two packets of 1480 + 20 (the outer protocol is IPv4, after all),
> > what is the problem?  Are people worried about the Segmentation and
> > Reassembly workload on the BR or something?
> 
> Oooh, yes, I definitely don't what the BR to do any kind of re-assembly.
> That'll limit the number of platforms that can be used for 6RD severely, a
> lot of the good price/performance ones are stateless and do not do re-
> assembly. Strictly speaking, having them to MSS adjust is a performance
> problem there already. Think Cisco 7600, hardware forwarding is good, CPU
> resources to manipulate packets is limited.
> 
> --
> Mikael Abrahamsson    email: swmike at swm.pp.se