Re: [TLS] TLS renegotiation issue

[Nicolas Williams <[e-mail]>]

Note, there's a way to implement Eric's proposal as-is and in an
SSPI/GGF-GSS-API way that doesn't suck, but which does require a
layering violation, though one that I could live with:

 - client and server apps both extract the 'tls-unique-for-telnet'
   channel binding from the outer connection, prefix that with the
   RFC5056-required prefix, and pass that to the init/accept_sec_context
   functions.

 - The TLS init/accept_sec_context implementation checks to see if the
   channel binding input starts with "tls-unique-for-telnet:" (the
   RFC5056-required prefix when using that channel binding type), and,
   if it does, then split the rest of the channel binding input
   accordingly and use the two halfs as per-Eric's proposal.

I can live with that.  I'd like to have a generic channel binding
facility in TLS.  I'd like to not have to implement such a layering
violation.

Nico
--