Re: [TLS] something something certificate --- boiling a small lake

[Nico Williams <[e-mail]>]

BTW, thanks for the something-something-certificate work.  Looking at
the I-D, draft-bdc-something-something-certificate-04, I see there's no
way to send the certificate chain on.  I understand the motivation
(compression), but it really would be best to send on the full chain
sent by the client.  More on this below:

On Fri, Jun 19, 2020 at 12:50:17PM -0400, Michael Richardson wrote:
> Thus, a single header isn't enough, although there could be some degeneration
> that results in a single header.  We need a few variables to update.
> 
> I think we have a choice between:

HTTP/2 and, I imagine, /3, does header compression of the right sort
here.

If the backend is HTTP/1, then... that sucks, but maybe this is a good
way to encourage migration to /2 or /3.  That said, I think we can
compress in the /1 case by first sending the certificate (and chain,
please) and in subsequent requests in the same connection sending only a
hash of the certificate.  This would force a /1 backend to keep the kind
of state that a /2 backend would for header compression.

> 1) sending the state (possibly a few kb) on every transaction, which keeps
>    the protocol stateless.  We could explore ways to encode it: CDDL+CBOR
>    seems like a good thing.  TLS structures are another obvious choice, but
>    that's a detail.

There's no need.  HTTP/2 already does header compression.

> 2) assuming that state will be maintained by both ends, and simply updating
>    the state appropriately.   When it comes to this, I think of the
>    HTTP PATCH methods, but I'm not sure I mean this literally.

Can TLS let a client authenticate multiple times?

> Alternately, the TLS front-end could maintain a RESTful interface on a
> per-connection basis that the back-end could interrogate.  The header
> would just provide the right reference to that.  The RESTful interface
> could even be pushed/updated into some other CPU on the TLS terminator.

Yes, this would also work.  In this case Client-Cert: would carry just a
URI.  This is nice because the backend can validate that the origin of
that URL is one it trusts.

Nico
--