Re: [patch] delayed acks decrease performances in congestioned networks

Jeffrey B. Siegal (jbs@quiotix.com)
Wed, 18 Nov 1998 19:44:00 -0800


>- - People will enable it which will lead to more congestion because of

>more packets.

Like NAGLE, delayed ack is a bet that may or may not pay off. NAGLE is
betting that more forward data will be received and can be piggybacked
with the delayed data, reducing network traffic. Similarly, delayed ack
is a bet that there will be a response generated for the data received,
with which the ack can be piggybacked. In both cases, there are
applications where because of the nature of the data being exchanged,
these assumptions are clearly false, and delaying the data hurts
performance without reducing network traffic.

The application knows best. There is no good reason to discard this
information simply because some people might misuse the feature. If an
application *really* wants to be anti-social, it can use UDP and build
its own protocol without delayed ack or any congestion control.

-
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/