At 17:09 23.11.00, Lloyd Wood wrote:
>On Thu, 23 Nov 2000, Reiner Ludwig wrote:
> > A compliant TCP has no other chance but to go into go-back-N after a
> > spurious timeout, i.e., a spurious timeout forces TCP into the go-back-N.
>
>even with use of SACK and knowledge of the receiver window state? (Did
>I miss a timeout causing the SACK scoreboard to be discarded?)
Yes, even with SACK. Note, that the first thing(s) a sender sees after a
spurious timeout are the returning ACKs that were triggered by the
*original* segments arriving at the receiver. Those ACKs will not carry any
SACK info.
Sure, *after* the go-back-N has been performed, the sender will then find
out what went wrong when it receives D-SACK info in the ACKs that are
triggered by the duplicate segments sent during the go-back-N. But at that
point it's too late; an entire flight has already been resent unnecessarily.
///Reiner
This archive was generated by hypermail 2b29 : Mon Jan 28 2002 - 09:12:25 EST