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

Re: Evolution of the IP model - ICMP and MTUs



Iljitsch van Beijnum  - Le 8/18/08 11:17 AM :
On 18 aug 2008, at 8:45, Rémi Després wrote:

IMU,

IMU?

since it shows that, at high data rates, IPv4 fragmentation can lead to undetected data corruption at the IP layer, it implies that fragmentation SHOULD be discarded from an updated IPv4 service model (the DF bit MUST be set in all packets).

What about actually FIXING the problem.

I remember that when I first read about IPv6 (many a moon ago) I noticed that IP packets had an unfragmentable and fragmentable parts. So I thought "excellent, they got it right, the port numbers are now in all fragments!" Unfortunately, that wasn't the case.

However, we could come up with a new fragment header for both IPv6 and IPv4 that DOES have all the information NATs and firewalls need in the fragment header, as well as a larger ID field. This would of course take significant time to get deployed, but considering that we've been limping along with broken PTMUD for a decade and a half having a GOOD solution may be worth the wait.


Fixing the problem for IPv6 may be worth the pain, but fixing it for IPv4 (the only subject of my comment) would IMO be counterproductive.

Do you know applications other than NFS on UDP that, needing to transmit longer than than 1280 octet datagrams, impose fragmentation even in IPv6?

If there is none, the "fixing" could be:
- Always use NFS on TCP.
- In IPv6, firewalls that cannot perform their function properly on fragmented datagrams impose a 1280 octet MTU (and return ICMP error messages when discarding fragmented datagrams).

RD