Re: Secure Shell: Milestone Update.

[Bill Sommerfeld <sommerfeld%sun.com@localhost>]

On Thu, 2005-03-17 at 12:24, Jon Bright wrote:
> > Well, I have an independent implementation of much of it which I'd be
> 
> I have an independent client implementation.  I'd be happy to do and 
> document interop testing.  What level of documentation is needed?

See the "Implementation reports" filed in:

http://www.ietf.org/IESG/implementation.html

Some of these are very informal, but as with everything else in the IETF
there has been increasing formality over time.  There's some history
here; see, in particular,
http://www.iab.org/appeals/snmpv2smi-appeal.html, which caused RFC2026
to specifically call for feature-by-feature testing.

In our case, the protocol is large and complex and has a large number of
optional and occasionally-used features.  I think the way to go would
be:

 1) go through the documents, coming up with a list of distinct
features.
Post this feature list to the WG for review.

 2) based on this list, send out an implementation survey asking, for
each feature:
	- have you implemented it?
	- has it successfully interoperated with another implementation
		(and, if so, who?)

Separate client and server surveys may be appropriate.

 3) collect survey responses

 4) figure out which features don't have coverage, and either prod
implementors to fill in the gap or decide the gap isn't a problem    
(stripping/deprecating the feature).

Some working groups have held interoperability tests, either
at/concurrent with IETF meetings, independently, or at other events. 
Given that many of our implementors have very limited travel/expense
budgets, pairwise interoperability testing over the internet by mutual
arrangement may be the best bet.