> On Sep 5, 2018, at 12:30 PM, Phillip Hallam-Baker <phill@xxxxxxxxxxxxxxx> wrote: > > Meanwhile we have had DANE throw in a service description attribute record but only for one transport protocol and only for one trust model and even though it has seen precisely zero adoption in the WebPKI world, it is apparently the only security policy approach we are permitted to consider. By WebPKI world, I assume you mean web browsers using HTTPS? DANE *has* seen adoption in SMTP (~314,000 domains presently), with implementations in multiple MTAs: Postfix, Exim, MailChannels, Halon, PowerMTA (beta), Cisco ESA (beta), ... For the Web, the main obstacle is last-mile DNSSEC issues, which may get easier as DNS over TLS and DNS over HTTP offerings from Cloudflare and the like eliminate bypass the CPE DNS breakage. The remaining obstacle is purportedly latency, which may be addressed with tls-dnssec-chain, if we manage to get adequate downgrade-protection into the spec. Finally, DNSSEC adoption is still light, at ~9.5 million domains, highly concentrated in Northern Europe, the USA and Brazil. The actual payload of DANE TLSA record has not been a real barrier to adoption. The PKIX-TA(0) certificate usage is a reasonable candidate for hardening HTTPS against rogue CAs. True, DANE is not "service discovery". If a new service discovery protocol that subsumed DANE became popular, and vended both service location and security policy, that'd be fine. It is an interesting problem to work on. -- Viktor.
