On Sun, Apr 26, 2015 at 11:48 PM, 🔓Dan Wing <[email protected]> wrote: > > On 26-Apr-2015 08:41 pm, Watson Ladd <[email protected]> wrote:
>> If what we end up with is doing the same crypto operations as >> DNSCrypt, but with the extra complexity of managing connections >> (opening, closing, resuming, etc), what is the advantage? > > TLS has an engaged community. The TLS protocol and its implementations have > been attacked and improved through the years -- "better the devil you know." As the person who broke SSL/1.0, I find that a rather weak argument. SSL could have been fixed much earlier if Andressen had handled things differently. If they had been willing to circulate the design prior to shipping the bits we could have gone straight to TLS/1.0 in 1995. Simon Spero and myself were both willing to work on it at no cost to Netscape. A lot of the difficulty since has been working within the limitations of some bad design decisions taken before anyone with crypto experience could comment on the design. I am not proposing to replace the key agreement part of TLS which is the complex part. All I propose to change is the framing part which TLS does in the opposite fashion to current best practice in any case. In effect I am proposing a different DTLS variation that is better suited to DNS traffic patterns by being completely stateless. _______________________________________________ dns-privacy mailing list [email protected] https://www.ietf.org/mailman/listinfo/dns-privacy
