* Zachary Amsden said:
> Chris Wedgwood wrote:
> >=20
> > > I suggest that tcp_sack be turned off by default unless there is
> > > some way to detect and work around broken implementations on the
> > > other of a connection.
> >=20
> > Do you know for sure it broken and not something else going on? I
> > have SACK enabled on machines that make hundred if not thousands of
> > connections to both solaris and windows machines every day... if
> > there are problems, I've not seen them -- but then again, I'm
> > certainly not looking either.
>=20
> But how lossy are those connections? In any case, the occasional noosed =
by
> SACK connection on the server side looks just like a hung or disconnected
> client. It when you are the client that you'll notice trouble.
That's basically what happens here - people report that their connection
seems to be broken, or just that they cannot load, e.g., one WWW page in one
run - two or three reloads are needed. Also, getting ANYTHING long with POP3
or IMAP also tends to break. Personally, I think that not only SACK is to be
blamed for those problems, but also the window scaling option - when I
turned BOTH of them off, the problems disappeared. When only SACK was off,
the problems were much less frequent, but nevertheless still existent.
marek
--nFreZHaLTZJo0R7j
Content-Type: application/pgp-signature
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.0 (GNU/Linux)
Comment: For info see http://www.gnupg.org
iEYEAREBAAYFAjhRnxgACgkQq3909GIf5urFZwCfaaYuYOtQAFSI4rsWpO+zTJL4
LfgAn1E5rCWsVAy7bJbfaoxhS4HqOauP
=Mb7v
-----END PGP SIGNATURE-----
--nFreZHaLTZJo0R7j--
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/