On Nov 21, 2017, at 12:42, Olaf Bergmann <[email protected]> wrote: > > Ludwig Seitz <[email protected]> writes: > >> I have a vague memory of a DICE draft for doing the DTLS handshake >> over CoAP a long time ago: >> >> https://datatracker.ietf.org/doc/html/draft-schmertmann-dice-codtls-00 >> >> Can the original authors tell us why they didn't go further with that >> approach? > > The main reason was that it did not fit into the official DICE work > program. Meanwhile, the student who did most of the work (including a > working prototype[1] which is also available as rubby gem) has graduated > and left the university. > > But there is no technical reason for not taking this up again. > > [1] https://github.com/SmallLars/codtls
This work was a successful early proof of concept and has thus fully achieved its objective. One more reason that we haven’t picked this up again is that the original work was for an older chip, the MC1322x, and contained code to keep connection state in the Flash of that chip that probably would need some porting. At the time, the TLS WG also didn’t seem particularly receptive to new work: codtls is not just a compressed DTLS, as it performed the MAC operations for the Finished messages on the actual messages exchanged, not on the original uncompressed messages that never are reified. Essentially, a large part of the complexity of the solution is in trying to maintain the entire set of TLS exchanges, and then the MAC doesn’t quite. (This is a fundamental problem any TLS emulation will have, and it would be good to get the envelope of what works and what doesn’t for this specified.) With much more tailored solutions around the corner (e.g., EDHOC), we didn’t see a strong technical motivation to take this up again either. But we sure can if there is a political reason instead. Grüße, Carsten _______________________________________________ Ace mailing list [email protected] https://www.ietf.org/mailman/listinfo/ace
