Hi,
So far I've received only a single response to my question about
strange nagle behavior under Linux. Below are three more pointed
questions on this topic. Feel free to direct me to an old thread on
these lists or elsewhere if these questions have already been
addressed. I've had a look through the archives and didn't see anything
relevant.
First, the Solaris machines I tested did not exhibit this behavior.
Apparently, Solaris (and some other commercial Unix flavors, I'm told)
apply the nagle rule only when there are < MSS bytes of outgoing data
in a send call. On the face of it, this does seem like a more
reasonable approach to the problem the nagle rule was designed to
correct. So, why does Linux delay fragmented packets instead of
small, single packet sends? And why are the delays so long with the
2.0 kernels?
Second, in the 2.0 kernel, the half-second delays waiting for ACKs on
*all* fragments, even if they are 99% full, seems like a big mistake.
Is there any good reason to do this?
Third, why the change in MSS from 1460 to 1448 in the 2.2 kernels?
I'm signing off of these lists (linux-net and linux-kernel) so please
send responses directly to me.
Thanks,
Chance
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
