Performance Tradeoffs of TCP Selective Acknowledgment
Apr 10, 2008, 04:30 (0 Talkback[s])
(Other stories by Patrick McManus)
[ Thanks to LinucksGirl for this link.
"Mixed in with the usual Linux development chatter on the
Internet over the last few months has been a significant discussion
of Linux's TCP SACK (Selective Acknowledgment) implementation.
These comments have generally focused on the performance of the TCP
stack when processing certain SACK events, and some people have
hinted at the presence of a security exposure.
"I was intrigued by the discussion, but it seemed to lack hard
data. What specific conditions were they talking about? Is this a
minor performance nit, or is it an outright server
denial-of-service (DoS) opportunity...?"