On Tue, May 13, 2014 at 3:41 AM, Dmitry Belyavsky <[email protected]> wrote: > Here are my ideas about "strict" behaviour of the TLS client: > > [...]
IMO log checking should be mostly asynchronous. This means that CT couldn't detect MITMing CAs in real-time, but that's OK because CT is about CA reputation, and all we care about is that when we detect lying CAs we can report them. This means that clients mostly should only reject TLS connections when there's no evidence of server certs being logged in an acceptable log. It also means we need some mechanism for reporting lying CAs. Thus far reporting via blogs, news outlets, and trust anchor set managers, has been good enough, and I'm not sure that we need anything more than any async notification to users (and upstreams) by browsers (and other clients). There are privacy protection considerations in reporting lying CAs: report the name about which they lied, just the log entry(ies) that are missing, or just an assertion that they lied? Again, as I see it CT is purely a reputation-based system. Reputation is established (and blemished/destroyed) asynchronously. This is important: we cannot be adding too much latency to TLS. By doing log checking asynchronously we won't be. Nico -- _______________________________________________ Trans mailing list [email protected] https://www.ietf.org/mailman/listinfo/trans
