The RFCs are not clear on this issue, but the basic congestion avoidance
principle is to only reduce once for a burst of losses within one window.
On Wed, 18 Nov 1998, Tang Oo wrote:
> I am not clear about the behavior of TCP SACK. If an ACK indicates multiple
> losses of segments, the congestion window should be cut by half, or cut by
> multiple halves? Is the fast retransmit and fast recovery still appliable to
> SACK?
>
This archive was generated by hypermail 2b29 : Mon Feb 14 2000 - 16:14:51 EST