ietf-nntp TLS and AUTHINFO interaction

[Jeffrey M. Vinocur]

On Mon, 17 Mar 2003, Russ Allbery wrote:

> Jeffrey M Vinocur <jeff at litech.org> writes:
> 
> > I was planning on lumping the already-authenticated state in with the
> > already-established-TLS state; in both cases STARTTLS would not appear
> > in list extensions, the client would be expected to know not to try it,
> > and any attempt to try it would be met with 500.  The two cases seem
> > very similar to me.
> 
> Hm... is it kosher to have LIST EXTENSIONS change after authentication?

You got me.  Seems to be what all the other protocols do, though.

Also, doesn't it *have* to change so that we can list e.g. SASL PLAIN 
which would not be acceptable on an insecure link?


> But if we could avoid making the client reissue LIST EXTENSIONS after
> authentication, I think I'd prefer it.  

I can't say I like any required use of LIST EXTENSIONS at all.  
(Particularly not for the convoluted way the forthcoming AUTHINFO USER 
text has it, but that's a different issue.)  But it seems that the de
rigueur thing is to compare the list of extensions before and after to 
prevent forced cipher downgrades, and if they have to do it anyway, I 
certainly feel like we may as well not list the things we don't want the 
client to try.


-- 
Jeffrey M. Vinocur
jeff at litech.org