Clear-Text Patch? was: Re: OpenSSH 5.1: call for testing

[rapier]

Dag-Erling Smørgrav wrote:
> Ben Lindstrom <mouring at eviladmin.org> writes:
>> [...]  IIRC "none" cipher was one of the first things removed from
>> OpenSSH back when it was forked.
> 
> Last I checked, it's still there; you just need to add "none" to the
> list of accepted ciphers in myproposal.h.

The problem is that just adding 'none' back pushes all interaction into 
the clear which is, IMNSHO, a really bad idea. While I personally think 
the none-cipher switching is a reasonable compromise between security 
and performance I fully understand and support the OpenSSH decision to 
not include it (even if I don't completely agree with it). There is, one 
can argue, a certain expectation that all SSH communications will occur 
with full encryption and the none switch, without a doubt, opens the 
user to additional risks. I've done my best to minimize those risks but 
they're still there.

So I'll continue to make the none switch patch available to users who 
are willing to assume those risks. What would be interesting would be if 
OpenSSH would be willing to support the idea of cipher switching in 
general - so that a user could move from one cipher to another 
mid-stream. There would be no need to include the none cipher in that 
list of allowed ciphers (but obviously, I'd like to build on top of an 
approved mechanism).

Either way, this is OpenSSH so the dev team gets to make these decisions 
- and live with those decisions. This is a heavy responsibility and why 
I fully support the decisions they make regarding their own software.

The main question I have is if they've increase the number of 
outstanding requests in SFTP to match the increased receive window size 
introduced in 4.7.