Re: [PATCH] Using TCP_NODELAY unconditionally

[Rick Jones <rick_jones2%hp.com@localhost>]

> Rick Jones makes the argument, on the OpenSSH list, that forwarding of
> other protocols should be done with Nagle turned off, that the
> forwarded traffic's characteristics would thus mirror the
> characteristics of the original traffic. I agree. I just don't know
> whether it would be problematic to run interactive character-based
> sessions without Nagle.
> 
> If SSH runs best with Nagle turned off then the draft RFCs should
> state it, possibly  with "SHOULD" language.

However, I would hope that any such language woud include something to
the effect that nagle in and of itself is not bad, and that disabling it
in ssh is done when acting as a pipe (man in the middle) and so it is
simply preserving (or at least trying to) the behaviour of the
applications on either end, and that any implementation of ssh SHOULD
NOT gratuitously "break-up" the data it recieves as a pipe etc etc
etc...

rick jones
normally a rather zealous opponent of arbitrarily setting TCP_NODELAY

-- 
Wisdom Teeth are impacted, people are affected by the effects of events.
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to raj in cup.hp.com  but NOT BOTH...