gson%gson.org@localhost (Andreas Gustafsson) writes: > Steven Bellovin wrote: >> > I wonder if it would make sense for the kernel to keep a list of >> > networks to disable SACK on, or look at TTL values to determine that. >> >> In theory, one should never have to disable it explicitly; its use >> is negotiated. But there's no particular harm to using if if both >> sides support it; it's just that it's not likely to help much. > > In my experience, on NetBSD, SACK does cause harm, by breaking > TCP Fast Retransmit. See the discussion on tech-net in 2008: > > http://mail-index.netbsd.org/tech-net/2008/01/28/msg000168.html So probably our SACK implementation should get fixed, unless this is really intrinsic in the spec (which seems hard to believe).
Attachment:
pgpm20Hh8C69I.pgp
Description: PGP signature