openssh unix dev - Jan 2010 - moving X11 portforwarding out into a "plugin" framework

I think everyone will admit that X11 forwarding has been an incredible feature
in [open]ssh.  X11 is not the only local->remote protocol that might be
useful
across an SSH session however.

But having to hack the code for new protocols as they come around seems silly.  
Wouldn't it be more useful to be able to describe a protocol that needs 
forwarding and some configuration that might need doing on each end outside of 
the source code?

Then as new applications come along that could take advantage of port forward-
ing can "plug in" to this system and get their protocols forwarded
across X
without having to hack the openssh code and push it upstream, etc.

I would imagine on the local side, the ssh client would need to be able to 
forward into unix and ipv4/6 sockets and on the remote side one would need to
be able to possibly create sockets and/or set environment variables, etc.

Thots?

b.

Brian J. Murrell wrote:
> X11 is not the only local->remote protocol that might be useful
> across an SSH session however.
Besides TCP it's the only one specified by the standards though.


//Peter

On Fri, 22 Jan 2010, Brian J. Murrell wrote:
> I think everyone will admit that X11 forwarding has been an incredible
> feature in [open]ssh. X11 is not the only local->remote protocol that
> might be useful across an SSH session however.
>
> But having to hack the code for new protocols as they come around
> seems silly. Wouldn't it be more useful to be able to describe a
> protocol that needs forwarding and some configuration that might need
> doing on each end outside of the source code?
>
> Then as new applications come along that could take advantage of port
> forward- ing can "plug in" to this system and get their protocols
> forwarded across X without having to hack the openssh code and push it
> upstream, etc.
>
> I would imagine on the local side, the ssh client would need to be
> able to forward into unix and ipv4/6 sockets and on the remote side
> one would need to be able to possibly create sockets and/or set
> environment variables, etc.
You should be able to do most of what you want using a Subsystem (see
sshd_config) and a helper program on the client side.

-d

Alex Bligh wrote:
> Really? Is -w / tun device standardized? The only docs I could
> find were in the source.
tun is a vendor extension:
http://www.openbsd.org/cgi-bin/cvsweb/src/usr.bin/ssh/PROTOCOL?rev=HEAD

-- 
Darren Tucker (dtucker at zip.com.au)
GPG key 8FF4FA69 / D9A3 86E9 7EEE AF4B B2D4  37C9 C982 80C7 8FF4 FA69
     Good judgement comes with experience. Unfortunately, the experience
usually comes from bad judgement.