[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: address pair exploration, flooding and state loss



Hi Iljitsch,


El 01/06/2005, a las 12:31, Iljitsch van Beijnum escribió:

On 31-mei-2005, at 23:03, Erik Nordmark wrote:

Another way to do this would be the "ICMP bad checksum" message that I suggested a few days ago.


would this be a mandatory change (i.e. required to make the SHIm work) or just an optimization?


I mean, i don't think that it would be a good idea to require mandatory changes on transport layers, even if they only involve TCP, UDP and SCTP...

Regards, marcelo


But as I responded, such an approach only makes sense if we are going to modify all the protocols which run over IP to generate such errors. And assuming that an implementation can do this is problematic due to "raw socket" APIs which allow there to be any number of IP protocols implemented unbeknown to the provider of the IP and shim6 layers.

AFAIK, applications that use raw sockets for actual user traffic are extremely rare. (I can't think of any, not counting IP tunnels.) Typically, applications use UDP for "strange" protocols. So if we modify TCP, UDP and SCTP we should be out of the woods here. And if there really are applications that don't use these transports but are interested in recovering quickly from shim state loss, it shouldn't be too hard for them to implement the "icmp bad transport checksum".