Ben,
My thinking is that:
1. There is concern, generally, about the size of certificates (really
about the size of handshakes).
That concern varies a lot, depending on context. I don't agree that this
is a universal, major
concern. Look at JSON; it's use of XML encoding suggests no concern
about bloat in a web context.
2. Embedding SCTs in certs leads to more bloat than either OCSP
stapling or the TLS extension because the SCTs cannot be updated, and
so more are required (see Chrome's EV/CT policy).
I am not familiar with Chrome's EV/CT policy, and I think you argued
that what Google has chosen to
do is out of scope, in the context of a prior argument in TRANS. Anyway,
absent a spec for TLS client
behavior, one cannot discuss the need to update SCTs, right?
3. Embedding SCTs in certs runs the risk that the SCTs will become
invalid before the cert does.
until we have a precise description of how a client will deal with an
invalid SCT, we can't
really evaluate the implications of this potential mismatch. Also, what
causes an SCT to become
invalid?.
4. Certs are not generally updated before expiry.
agreed.
5. Stapled OCSP requires the CA, the webserver and the CT logs to be
actively involved.
agreed. which gets back to my question from several months ago as to why
we have 3 ways
of signalling SCT info.
6. The TLS extension only requires the webserver and the CT logs to be
actively involved.
agreed, if the web serer operator submits the cert to the log.
7. All three mechanisms can be made essentially automatic from the
webservers POV.
This point, in conjunction with #5, sounds like you're trying to have it
both ways :-).
Lots of things can be made automatic, with the right support software.
In the IETF
we often try (though we do not always succeed) to limit the number of
mandatory to
support ways to do the same thing. That's why I questioned the
specification of OCSP
delivery of SCT data, plus pre-certs plus TLS handshake delivery.
So, I would expect in the long run that the TLS extension would be the
favoured option.
Maybe. But, as I noted, if CAs have already gone to the trouble of
embedding SCT data in
certs, and if this is free for web site operators, there seems to be
more inertia than
motivation to change, from the perspective of operators.
Steve
_______________________________________________
Trans mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/trans
