Vishwas Manral wrote: > Hi, > > I had raised this earlier in December 2005 > http://www1.ietf.org/mail-archive/web/ipv6/current/msg06020.html . > > The definition of IPv6 fragment is not clear. Infact different > protocols assume it differently, example are IPsec and SIIT. I do not > have implementations to play with, however I feel we need to close > this issue along with other related issues, like tiny fragments. The > discussion there was:
Well, SIIT doesn't provide definition for the IPv6 fragment. It simply adds some special cases to handling the fragment header under SIIT conditions. > > " > Vishwas Manral wrote: > > I have a doubt regarding the fragment header. Why do we need the M > flag in the "fragment header" at all for IPv6? Having the fragment > header itself would tell it's a fragment and would distinguish between > the first fragment and a non-fragment. How do you tell the terminating fragment from a non-terminating one without an M bit. > > In IPv4 we did not have a fragment header, so the M flag was logical > to have for distinguishing the first and a non fragment. > > That said; how should the case where we have the fragment header and > both the Fragment Offset and the M flag is 0 be treated?" Handle it as a self-terminating fragment, ie. you reassembly queue is 1 fragment long. This simply follows the spec as written. -vlad -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
