On Fri, Oct 01, 2004 at 10:28:00AM -0700, Bill Studenmund wrote:
> On Fri, Oct 01, 2004 at 10:05:44AM +1000, Daniel Carosone wrote:
> > 
> > It's hard to say exactly without looking at it, but the lack of *any*
> > ack's is a little confusing there. You should see at least one more
> > re-ack when the out-of-order segment comes in. (just a little green
> > tick on the line)
> 
> What have we changed with this stuff since 1.6? I have a MacOS X laptop 
> which opens tcp connections to a NetBSD box at work. Right after we 
> upgraded it to 2.0_BETA, I noticed tcp connections would stall. Downloads 
> gave like 10% or 5% of the link capacity. Oh, this is a WAN connection.

We haven't changed much, but MacOS X has a persistent bug in this
area, the "stretch ACK bug", which causes very similar symptoms when
talking to non-local peers that don't do path MTU.  Do you perhaps
have path MTU turned off for some reason?

Generally, our stack doesn't implement some stuff that it should
(but which is optional), such as SACK, rate-halving, etc., but the
MacOS X stack is *really* buggy, especially when talking to peers that
do not implement exactly the same set of optional stuff that it does.
The combination is particularly prone to failure...!

-- 
 Thor Lancelot Simon	                                      tls@rek.tjls.com
   But as he knew no bad language, he had called him all the names of common
 objects that he could think of, and had screamed: "You lamp!  You towel!  You
 plate!" and so on.              --Sigmund Freud