Re: [TLS] 0RTT?

[[e-mail] (Martin Rex)]

Watson Ladd wrote:
> Martin Rex <[e-mail]> wrote:
> > Watson Ladd wrote:
> >>
> >> I think I came up with a decent idea for handling 0RTT handshakes,
> >> namely making resumption 0RTT, and doing a side rekeying.
> >
> >
> > These idea(s) seem to based on a number of extremely optimistic
> > assumptions that have zero chance to get airborne in many usage
> > scenarios.
> >
> > The "regular" TLS session resumption means that client announces
> > a TLS session ID in ClientHello and server looks up that session ID
> > in its own cache for a match.  If there is no match, a regular
> > TLS full handshake proceeds.
> >
> > 0RTT is a no-go for a huge number of existing apps which do not
> > perform any fancy reconnect fallbacks at the application level
> > (closing the previous connection, opening a new connection and
> > starting a new handshake from scratch while requesting different
> > handshake characteristics this time).  For the general case
> > a 1RTT resumption is required, which can successfully perform
> > a full handshake in case that resumption is not possible,
> > without having to go through a connection/handshake failure.
> 
> I think we can handle this by thinking about fallback options more
> carefully. Certainly a resend can be done at the implementation level
> with the new handshake: this will cost latency, but it can be done.

For 0RTT this option does not exist.  A multi-megabyte HTTP POST on top
of a 0RTT TLS connection would waste lots of resources before even
realizing that there is a problem and the bandwith was completely wasted.

For other TLS extensions, a possibility to restart/reset the ongoing
handshake could be defined so that TLS implementations could adopt
and transparently provide it.

But look at what gets pushed here in TLS WG instead:
  http://tools.ietf.org/html/draft-ietf-tls-downgrade-scsv-00

which is a bogus proposal to panic and close the network connection
on the server side and dump a genuine TLS problem entirely on the
application caller of the clients TLS implementation.

-Martin