[Cryptech Core] peering into the dusk

Linus Nordberg linus at nordberg.se
Sat Sep 27 10:50:58 UTC 2014


Randy Bush <randy at psg.com> wrote
Fri, 26 Sep 2014 09:06:28 +0900:

| if this makes sense, perhaps jakob will lead the dnssec requirements
| discussion, and rob rpki and bgpsec.  and i presume linus will chime
| in with tor, which is good.

For signing consensuses (once per hour) we'll need RSA (with at least
2048-bit keys) and SHA-1 for now. I'm going to try to figure out for how
long this will be enough and when Ed25519 + SHA-512 is what we want,
too. I'll follow up on tech at .


Another early adopter I'd like to ask you about your thoughts on is
signing in Certificate Transparency (RFC6962), i.e. signing of STH's and
SCT's. The spec says RSA (2048-bit or longer keys) with SHA-256 or ECDSA
with NIST P-256. If we want to serve Google, ECDSA is needed.

Signing of STH's typically happens once per hour or less. I estimate
signing of SCT's to happen "some hundreds of times per hour" and that
"once per second" at peak is going to be just fine.

Thoughts on CT as an early case?



More information about the Core mailing list