Re: [saag] post-X509 cryptographic identities

[Stephen Farrell <[e-mail]>]

Hiya,

As a quote of mine seems to have started this...

On 11/02/2020 10:55, Michael Richardson wrote:
> 
> Stephen Farrell <[e-mail]> wrote in a IETF-last call thread:
>     > Anyway overall I take this as more evidence that
>     > x.509-based pki has outlived it's useful lifetime.
>     > Given the webpki needs CT (which it totally does)
>     > and now maybe novel revocation mechanisms like this,
>     > (as well as soon-to-be PQ schemes if we believe
>     > what people tell us), I'd argue it may well be time
>     > to try see if there's any consensus on a post-x.509
>     > direction towards which to head.
> 
> I agree with you strongly.

Great.

Unfortunately for me, I figure I disagree with the
direction of almost all the discussion in the rest
of this thread so far;-)

- I like DNSSEC, but at ~1% penetration in 20 years
it is not the answer, regardless of whether or not
it ought be the answer. (Even so, continued effort
to improve DNSSEC deployment by paying registrars
etc is good as it does mitigate real vulnerabilities
when deployed.)
- I like some of PHB's ideas wrt his meshy thing,
but I do not believe that a single all-encompassing
model like that that starts from better crypto will
fit all needs. I'd be happy to be wrong, but also
surprised.
- I like abstract discussions sometimes, but IIRC
I think I've seen those more abstract points made
various times over the years with them leading to
nothing much.
- While I probably got fed up with x.509 about the
turn of the century after a decade of writing code
for it, I recall that even before then I concluded
it had been a pretty bad error to lawyer-up any of
those discussions before we knew how to schlep the
bits around, so I'd run a mile if someone wanted
to start from there;-)

For me, I'd be interested in starting with a much
less ambitious goal, e.g. perhaps how to integrate
PQ algs into TLS and SSH without x.509, or maybe
whether we could do even better than ESNI inside
HTTPSSVC, or maybe if there's a way to unify smime
and pgp via new key packaging (that MUA folks would
accept). Or other relatively modest but still quite
hard things like those.

If some set of similar designs addressing a few
such concrete problems could be deployable alongside
the current stuff, and people started coding 'em up,
then we might be on to something.

Cheers,
S.