On 2012-01-05 02:57, RJ Atkinson wrote:
> Earlier, Mark Andrews wrote:
>> Atomic fragments > 1280 should not appear in the network.
>> Atomic fragments <= 1280 are a expected part of the IPv6 landscape.
>> For TCP they should be rare.
>> For UDP it depends on the protocol running on top of UDP.
>> PMTUD relying on PTB is just not reliable.
>
> To the best of my understanding, the above is correct.
The last point is IMHO the most significant. As far as I'm
concerned, PMTUD for IPv4 has *never* been reliable, and for
many years I clamped the MTU on my laptop at 576 to avoid
constant connectivity failures while on travel. That seems
to have become unnecessary in recent years since 1500 became
almost universal as the link MTU (but PMTUD is still unreliable).
I see no reason to expect that PMTUD will be more reliable for
IPv6 than for IPv4. 1280 is reasonably safe today only because
1500 link MTU is widespread.
On 2012-01-05 05:19, Dan Wing wrote:
> We can clamp TCP MSS on our various translator devices.
However, this won't help with TCP implementations whose
MSS negotiation is broken. I haven't seen that failure
mode on translated paths, but I have seen it on 6to4 paths,
where one end wanted to reduce the MSS and the other end
was stuck on 1440.
Brian
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
