On Jul 10, 2013, at 12:21 , Ronald Bonica <rbon...@juniper.net> wrote:
> 
> If ICMPv6 PTB messages are unreliable, fragmentation breaks just as badly as 
> PMTUD.
> At the risk of going off-topic, please take a look at 
> draft-bonica-intarea-gre-mtu-02.

I did. I won't comment about that here, because I too don't want to drift off 
topic.

Extending my remarks: Mr. Templin is right to point out that tunnels over 
networks where ICMPv6 Packet Too Big errors are not reliably delivered from the 
egress to the ingress, and also where such errors are not delivered from 
destination hosts back to the egress, have no choice but to assume the PMTU 
from the ingress to the egress is 1280 octets, and they must fragment packets 
that would exceed that length when encapsulated in an IPv6 header chain. If 
they cannot fragment, than paths through the tunnel do not meet the minimum MTU 
requirement.

I agree with him that SEAL addresses this problem, by using a different 
mechanism to fragment that the RFC 2460 extension header we are proposing to 
deprecate, but RFC 2473 doesn't do that, and there is a procedural issue I am 
trying to surface here: if we're going to deprecate the fragmentation header, 
while at the same time recognizing that the Internet is unable to deliver 
ICMPv6 Packet Too Big errors reliably, then we are blowing a big hole in the 
standards by not promoting SEAL or something like it as a STANDARD replacement 
for GRE and RFC 2473 encapsulation.

All I'm saying is that if we're going to start deprecating fragment headers, 
then let's make sure we are deprecating all of the things that rely on them, 
and providing a standard replacement for each of them in turn.

[Insert here shopped version of classic Hyperbole And A Half Cartoon, saying 
"DEPRECATE ALL THE THINGS!"]


--
james woodyatt <j...@apple.com>
core os networking

--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to