Re: b4-0.9.0 available

[Konstantin Ryabitsev <konstantin@linuxfoundation.org>]

On Fri, Jun 24, 2022 at 12:05:35PM -0400, James Bottomley wrote:
> On Fri, 2022-06-24 at 11:52 -0400, Konstantin Ryabitsev wrote:
> >    - submitters will be required to register their ed25519 patch
> > attestation key with the submission endpoint and cryptographically
> > sign their patches (email roundtrip confirmation required to
> > register).
> 
> Can we please be less pejorative about crypto choices?  ed25519 is a
> fine curve, but lots of crypto HSMs which do elliptic curve don't do
> Edwards signatures because they're very different from standard ECDSA
> signatures.  Those of us who take security seriously tend to use HSMs
> for all our private keys and thus won't do ed25519 if the HSM we use
> can't do it.

This will be only required on the web submission endpoint, and only to begin
with. Shelling out to call gpg or openssh from a web process is a lot more
hairy than checking ed25519 signatures using PyNaCl, which is the only
reason behind this restriction.

> I was under the impression b4 used gpg on the backend, which can do a
> variety of elliptic curves and now, bonus, can use a laptop TPM as the
> HSM to protect the keys (which means pretty much everyone has no excuse
> not to do hardware key protection)? (TPMs are one of the crypto devices
> that can't do Edwards signatures).

If you're not planning to use the web submission endpoint (and I expect that
anyone already using git-send-email will not be interested in the web
endpoint), you're certainly free to continue to use gpg.

-K