Re: [TLS] analysis of wider impact of TLS1.3 replayabe data

[Bill Cox <[e-mail]>]

On Sun, Mar 13, 2016 at 11:23 AM, Ilari Liusvaara <[e-mail]>
wrote:

> TLS 1.3 1-RTT is just boring, unless you are trying to do something
> at least a bit screwy, like mix pure-PSK and client-auth.
>
> No such luck with 0-RTT. There is all sorts of cryptographic screwyness
> in there too (through getting rid of DH-0RTT should eliminate that).
>

This is just my personal opinion:

I second your opinion that as Ekr suggested, TLS 1.3 should drop DH-0RTT.
Only 0-RTT resume should be supported.  That way, we start off with the
same security as 1-RTT, simplifying analysis and reducing vulnerabilities.

IMO, 0-RTT is the most important new feature in TLS 1.3, though the current
spec giving us two different ways to shoot ourselves in the foot with 0-RTT
seems redundant.  While this is in no way related to my job within Google,
I do not see how a company like Google could justify switching from QUIC
crypto to TLS 1.3 without a 0-RTT capability.  Note that the job I actually
do at Google, working on Token Binding, is significantly threatened by
0-RTT (just like client-certs), so I have plenty of reason to be against
0-RTT.  I just feel the world needs 0-RTT.  Speed really _is_ that
important.

Finally, I feel that while the TLS 1.3 spec does a good job of trying to
scare away people from 0-RTT, it would be better if we also explained what
is required to use 0-RTT in a secure mode.  Perhaps this does not belong in
the spec, but at least a link to read about it should be, IMO.  It _can_ be
done, if the client and server are willing to pay the high cost required.

Bill