gnutls-cli fails to handshake with Exchange server that uses DES-CBC3-SHA cipher

[Ted Zlatanov]

On Sat, 31 Mar 2012 19:32:16 +0200 Nikos Mavrogiannopoulos <nmav at gnutls.org> wrote: 

NM> On 03/30/2012 02:02 PM, Ted Zlatanov wrote:
>> On Thu, 29 Mar 2012 20:22:31 -0400 Thomas Fitzsimmons <fitzsim at fitzsim.org> wrote: 
>> 
TF> Emacs allows overriding the default GnuTLS priority string using a
TF> variable (gnutls-algorithm-priority) so I set it to "performance" to
TF> work around this server-side issue.  In cases where Emacs would
TF> otherwise fail to connect to a server because of a weak ciphersuite
TF> maybe the UI should warn the user and ask them whether or not to
TF> proceed.  Anyway, thanks for analyzing the logs.
>> I don't think currently Emacs can distinguish this case from a normal
>> negotiation failure.  The best we can do is to generally suggest a
>> weaker priority string, which seems to be a bad idea.  Is there a way to
>> determine that this case has occurred?

NM> You cannot in general distinguish a negotiation with a broken server and
NM> negotiation failure. What (I think) browsers do is if negotiation fails
NM> they fallback to the most compatible mode (SSL 3.0 or so).

So you're suggesting to try a weaker (more compatible) priority string,
right?  We could do that per server name.  Considering we have just one
bug report on this and from a broken server, I'm not sure it's worth the
effort to automate the fallback.  In your experience, is this a
widespread problem worth addressing through code, or is it better as a
FAQ?

Thanks
Ted