[pkix] Client-side OCSP stapling? Re: Proposal for working on PKIX revocation open issues

[Anders Rundgren <[e-mail]>]

Since you want to do something in revocation I would like to
describe an existing potentially global PKI-using system that
maybe could be improved.

The EU e-passport system needs for crossborder-checking of biometrics
a pretty elaborate PKI scheme which among many things require
the parties to expose two public ports on the Internet; one for
the actual communication using HTTPS[1] and another for publishing
CRLs using HTTP.  This isn't rocket-science but it still requires
multiple FW settings and proxies.  If OCSP responses could be
stapled (TLS client cert auth is used), relying parties would only
have to open a single inbound port.  Cross-border reliability would
probably also be improved since the client (sender) wouldn't be able
to submit any data unless its OCSP is running (the PKIs are unique
per country).

TLS 1.3 and 2.0 are in the workings so the timing is right...


Anders

1] I might add that I believe HTTPS with client certificate auth
is a very poor choice for cross-border communication when each party
run their own PKI. Signed messages permit a multi-tier architecture
and quarantining of not yet trusted messages, greatly simplifying
operation.  BSI are experts on crypto, but n00bs on IT :-)