Re: [TLS] Need for S->C signaling

[Marsh Ray <marsh at extendedsubset.com>]

Dr Stephen Henson wrote:
> Just had a crazy idea for S->C signaling:
> 
> ServerHello contains the MAGIC value in the CipherSuite indicator: we know this
> can never happen by accident. The real ciphersuite indicator is placed
> elsewhere, e.g. part of session ID. Session ID definition slightly modified to
> permit 34 bytes instead of 32.

I like it.

For reference: http://tools.ietf.org/html/rfc2246
> struct {
>         ProtocolVersion server_version;
>         Random random;
>         SessionID session_id;
>         CipherSuite cipher_suite;
>         CompressionMethod compression_method;
> } ServerHello;


Is changing SessionID from
>   opaque SessionID<0..32>;

to 34 bytes any less-breaking of a change than simply changing the
length of the ServerHello to make it a proper field? Maybe, I dunno.

What if the magic cipher_suite value signals that the acutal
cipher_suite value is stored in the top two bytes of random?

If anything actually cares about the GMT, he can fill those in with his own.

- Marsh