Re: Invalid channel numbers

[Simon Tatham <anakin%pobox.com@localhost>]

Peter Gutmann <pgut001%cs.auckland.ac.nz@localhost> wrote:
> Yeah, I guess that would do.  The reason why I'd prefer not to close the
> connection outright is that there might he half a dozen data transfers
> currently in progress, and I'd prefer not to abort them all just because of
> one malformed packet on an unrelated channel.

The question is, though, why would you _get_ such a malformed
packet? The SSH data channel is required to be free of corruption
and data loss (and the MACs enforce this vigorously), so it's not as
if half of an SSH message is going to routinely disappear en route.
The only way it can happen is as a result of a pretty fundamental
bug in the SSH implementation at the far end.

Therefore, it isn't unreasonable to assume that an SSH
implementation which has sent you a packet that badly formed is in a
state of total internal confusion, and to terminate the connection
before it sends you any more plausible rubbish that you might
accidentally trust!

I can't imagine any situation in which this sort of error condition
would occur routinely, such that it would make sense to spend effort
on salvaging everything possible from the rest of the SSH
connection. It really ought to only ever occur when the client or
server is in development, in which case you weren't transferring any
important data through it anyway.
-- 
Simon Tatham         What do we want?        ROT13!
<anakin%pobox.com@localhost>   When do we want it?     ABJ!