> > The MUST here seems to be implying that the only way to get the server to > > recognize an SID is to choose the one it's expecting. Since SYN packets may > > be reordered, this is best-effort. > > Perhaps an extra sentence saying the passive opener MAY tolerate some > reordering would clarify things?
That's probably fine. This feels like one of those cases where we want to encourage implementations to be liberal in what they accept, as the realities of network conditions guarantee that even following the spec to the letter will not always result in clockwork operation. Should the MUST be a SHOULD instead? The protocol still works even if the active opener skips every other SID... They may simply not benefit from the cached session, depending on how liberal the passive opener is. Kyle
_______________________________________________ Tcpinc mailing list [email protected] https://www.ietf.org/mailman/listinfo/tcpinc
