[Standards-JIG] Re: proto-JEP: Proposed Stream Feature Improvements

[Dave Cridland]

On Tue Aug 22 17:18:33 2006, Dave Cridland wrote:
> There's a problem here, though, which is that the client's stream 
> start can be pipelined with the end of the TLS negotiation, so in 
> that case at least, you gain nothing.

The above is wrong - I clarified it with Tony Finch, who's done quite 
a bit of work looking at streamlining ESMTP connection startup. I may 
still have it slightly wrong, which'd still be my fault, not his.

It's only true for the case where the TLS session cache is hit, which 
would normally be for a reconnection.

For "cold start" TLS, as it were, the server speaks last.

What this means is that if the server automatically sends you the 
post-TLS features, this will save you a round-trip for "cold start", 
but it means there's nothing gained by hitting the TLS session cache.

The moral of the story is that servers need long-lived TLS session 
caches.

Dave.
-- 
Dave Cridland - mailto:dave at cridland.net - xmpp:dwd at jabber.org
  - acap://acap.dave.cridland.net/byowner/user/dwd/bookmarks/
  - http://dave.cridland.net/
Infotrope Polymer - ACAP, IMAP, ESMTP, and Lemonade