Re: [TLS] RFC-4366-bis and the unrecognized_name(112) alert

[Martin Rex <[e-mail]>]

Joseph Salowey wrote:
> 
> The discussion in 5246 indicates that the warning messages are
> unpredictable.  I don't think we can solve this problem in this document
> so we wither discourage the warning in this case or we just explain the
> unpredictability.  Below are two proposed text change to paragraph after
> the structure definition in section 3 in draft-ietf-tls-rfc4366-bis-08.
> Let me know which you prefer soon as we need to get this document
> finished, because other documents are waiting on it.  

I do agree that we're talking about an extension (SNI) and an alert
for an existing spec (rfc4366) and an existing installed base,
so describing the risk of sending a warning-level "unrecognized_name"
alert to clients is sensible and appreciated.

Still, I think we have to document what _clients_ SHOULD and SHOULD NOT
do when receiving a warning-level "unrecognized_name" alert.  This was,
and still is not clear from the specification, rather it is essentially
undefined  (it is _not_ implementation defined, because a server
can not reliably distinguish individual peer client implementations by
looking at the contents of the ClientHello handshake message it sent).


-Martin