Re: QUICker connection establishment with out-of-band validation tokens

[Ted Hardie <[e-mail]>]

Thanks for your rapid reply.  More in-line.

On Fri, Apr 12, 2019 at 12:21 PM Erik Sy <[e-mail]>
wrote:

> On 4/12/19 18:49, Ted Hardie wrote:
>
> The article proposes two different distribution mechanisms for
> out-of-band tokens. The distribution via DNS resolvers is certainly more
> difficult to realize as it is a cross-layer performance optimization.
> The second proposed distribution mechanism using QUIC connections to
> other hostnames is much simpler to implement. Thus, I suggest to
> consider this mechanism to be in scope for QUIC v1.
>
>
The proposal for an EXTERNAL_TOKEN frame has security properties and
performance properties that would still need considerable analysis.  If the
token is associated with a hostname which is already present in the
certificate for the TLS session, it's not clear why the token would need to
be presented on a new connection.  If it is not, then the client's trust
that it is okay to receive a token for example.com from mumblefratz.example
is based on very unclear principles.  While you have addressed why you
believe the client is not harmed by presenting a bad token, it is not clear
to me that this is true for servers nor that server's behavior toward those
clients presenting bad tokens will be consistent.

I must reiterate my statement that I don't think this should in scope for
v1 at all.

regards,

Ted