Re: [hybi] Websocket success rates and TLS extension.

[Justin Erenkrantz <[e-mail]>]

On Tue, Apr 6, 2010 at 1:11 AM, Greg Wilkins <[e-mail]> wrote:
> Because TLS negotiations represent round trips, they
> don't want extra round trips for framing negotiations.
> Hence they have proposed a TLS extension to allow
> protocol negotiation during TLS handshake.
>
>  http://tools.ietf.org/html/draft-agl-tls-nextprotoneg-00
>
> This would allow a TLS connection to be negotiated
> with an application protocol and further round trips
> avoided.

Sorry for coming back to this, but Mike's later posts reminded me of
this comment as a way to reduce round-trips.

This is interesting, but the ID doesn't really contain a useful
description of how it would be implemented - it punts that out of
scope and that's the more critical bit, I think.  I'm just not sure I
understand what the "selected_protocol" field would say at all.

Regardless of the vagueness of this ID, doing something to optimize
latency is important.  If this TLS extension makes it way into
OpenSSL, I'm sure httpd/mod_ssl could easily pick up on it.  Is an
implementation available or planned?

I don't know how hard it is for Java servers to support TLS extensions
- I'm sure you know, Greg.  =P  -- justin