Tomas Mraz pí¡e v St 08. 11. 2006 v 12:07 +0100:
> On Wed, 2006-11-08 at 10:55 +0000, Andrew Haley wrote:
> > Adam Tkac writes:
> > > I think, It's no argument to include rsh in next versions of
fc/rhel.
> > > OpenSSH could successfully substitute this component. SSH is more secure
> > > than rsh and has all features of rsh. Do you think anything else??
> >
> > High performance environments are one obvious example. I've used rsh
> > for high-load testing of systems, where you don't want encryption
> > overhead to get in the way.
>
> Yes, definitely. There are also other problems with scp/sftp protocols
> which lower their performance on fast networks. These are being remedied
> with the HPN patch (
http://www.psc.edu/networking/projects/hpn-ssh/) but
> upstream still refuses to accept the patch and we don't want to diverge
> from it so much. There is also a question if this patch couldn't cause
> incompatibility with unpatched OpenSSH clients/servers in some
> circumstances.
>
> So I'd say, keep the rsh for now. Or even better possibility would be to
> move it to Extras as Extras are NOT a second class citizen and RHEL can
> still include it.
>
> --
> Tomas Mraz
> No matter how far down the wrong road you've gone, turn back.
> Turkish proverb
>
You're truly right. This idea about high-performance systems doesn't
strike me. I'm going to look forward to this idea when upstream of
OpenSSH accepts HPN patch (or other improvement for this problem).
The acid test is this, run over a fast network:
$ tar cf - foo | rsh bar tar xfBp -
This should hit the bandwidth of the connection.
Andrew.