Iljitsch van Beijnum le (m/j/a) 8/20/08 10:40 AM:
On 20 aug 2008, at 9:59, Rémi Després wrote:DCCP seems to me the right direction for graceful transmission of datagrams that exceed path MTUs through filtering devices....the port numbers are still only in the first fragment,
Well, there remains a terminology ambiguity in the fragmentation DCCP option I propose to introduce. - Let's say that what applications want to send are "User Datagrams", and that what DCCP submits to the IP layer are "IP Datagrams". - Each User Datagram that is too large to fit in an PMTU-compatible IP Datagram is then split by the DCCP layer into several "User Datagram Fragments", each one being sent in a single PMTU compatible IP Datagram. - Since each of these IPDs is Path MTU compatible, it remains E2E as a single IP layer fragment. It therefore keeps its DCCP port numbers, independently of its being the first or another fragment of its User Datagram.
so filtering remains somewhat problematic and NATing very problematic.
Not in this case, which it the result that was looked for :-). Rémi Després