From mboxrd@z Thu Jan  1 00:00:00 1970
Date: Fri, 15 Sep 1995 13:22:07 -0400
From: forsyth@plan9.cs.york.ac.uk forsyth@plan9.cs.york.ac.uk
Subject: faster u9fs
Topicbox-Message-UUID: 247d8472-eac8-11e9-9e20-41e7f4b1d025
Message-ID: <19950915172207.1VAqg3WHLlwKpq0fii84AmtAI89SPWVLkhmWVs2Zqgo@z>

if your unix tcp/ip implementation has the TCP_NODELAY option,
try enabling it in u9fs:

	int one = 1;
	setsockopt(fd, IPPROTO_TCP, TCP_NODELAY, &one, sizeof(one));

you'll have to scour the heap of include files in /usr/include
to find the one with the TCP_NODELAY option in it.

i found it made a difference on our rs/6000 systems.

9P uses many small reply messages, which are subject to the delay.
(the system waits to see if more data turns up to make the
packet really worthwhile, but in this case it just makes it less responsive.)