> On Nov 20, 2019, at 8:14 PM, Manoj Nayak <[email protected]> wrote:
>
> Hello Fred,
>
> Yes, it is possible that the largest fragment received is *much* smaller than
> the PMTU. However, a survey of
> popular operating systems reveals that the largest fragment does reflect the
> PMTU.
That’s great if that’s the only place this ever runs...
> If we are really worried about this problem, the sender can ignore the ICMP
> message when the MTU is smaller than
> the “smallest believable value” (e.g., 1500 bytes).
I suppose that means you don’t “believe” in tunnels.
Seriously, all this work to get direct report from the endpoint of a packet
size that works
- using a protocol that we don’t rely on because it’s generally blocked
- providing a number that might work, if you run on “popular” systems
- all to then throw it out because it isn’t what you “expect”
> There was another query, if it is worth doing Lossless PMTUD for ipv4.
> IPv4 will be with us for a long time. Lossless PMTUD doesn’t take much
> effort. So why not …
I provided a partial list above.
Joe
_______________________________________________
Int-area mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/int-area
