On Wed, Jun 18, 2008 at 1:52 AM, David Miller [off-list ref] wrote:

It's pretty easy, depending upon the timestamp resolution, to
end up with the original transmit and the retransmit having
the same timetsamp value.  So this suggestion in the RFC is
I think not a complete solution.

I happen to have just (unintentionally) simulated this, i.e. a
situation where MD5 signed packets with SACKs were dropped by the
receiving host. The transmitting host keeps sending packets with SACKs
and they keep getting dropped: the connection stalled.

AGL

-- 
Adam Langley agl@imperialviolet.org http://www.imperialviolet.org