On 2012-01-10 06:08, Fred Baker wrote: > On Jan 6, 2012, at 8:26 PM, Fernando Gont wrote: > >> I just said that PLPMTU has a long convergence time > > Frankly, that sounds like an implementation issue. If we're willing to set > IW=10, it seems like one should be able to send the initial burst, or at > least the first retransmission, with messages of several sizes and see what > works. In other words, if the MSS is 9K (I wish), send a 9K, a 4K, a 1500 > byte IP datagram (eg 1440 byte payload), and 1280. If we get an Ack in the > subsequent RTT acknowledging 8192 bytes, we learned something, and if the > only Ack acknowledges 1280, we learned something.
Right. And the point Fernando queried in my previous message was simply trying to say "happy eyeballs doesn't do this" - it will leave the user hanging if TCP connects but PMTUD or MSS negotiation fails. Brian -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
