Re: [PATCH] Using TCP_NODELAY unconditionally

[Tobias Ringstrom <tori%ringstrom.mine.nu@localhost>]

On Tue, 22 Jan 2002, Nicolas Williams wrote:

> If interactive character-based sessions are the one place where Nagle
> definitely belongs and OpenSSH turns it off there, why not turn off
> Nagle altogether for all SSHv2 sessions?

This is moving away from the scope of ietf-ssh, and back into 
openssh-unix-dev...

I'm not familiar with the SSHv2 spec, but would it be possible, when
opening a new channel, to specify whether the channel should be nodelay or
not.  Is there a standard-compliant way to do this?  A backwards
compatible way?  If such a facility existed, the implementation could set
TCP_NODELAY if one or more channels were nodelay channels.

/Tobias