Re: VC formats

On Tue, Mar 19, 2024 at 8:20 PM Kim Hamilton <kimdhamilton@gmail.com> wrote:

> That comparison matrix is gold, thanks! Spice isn’t there but that did
> come after IIW.
>
> There’s a lot in there and it seems a bit overwhelming. However I think
> most of us are necessarily abstracting away from this level, assuming a
> multi-model/format/etc world, to focus on business value, other aspects of
> the ecosystem, etc.
>
> As that happens, I think communities like this can play an important role
> in facilitating understanding of impact of these differences on people.
> Exciting stuff ahead!
>

Kim,

I also want to make it clear that Gordian Envelope is somewhat at a
different layer than the other examples (SD-JWT, SD-CWT, ecdsa-sd, mDoc) in
that it is more generalized to be useful for any authenticated data, in
particular data at rest, and thus is not solely for credential data. Its
focus is more on data minimization, and can be used for health-care data,
AI foundation models, business data, other forms of PII, etc. Gordian
Envelope can be used for credentials as well, but right now there is no
funding to make it aligned with VCDM. Should be possible, but you lose some
of the privacy benefits that allows any holder (not just subject-holder,
but any holder) more choices for what to selectively disclose, or
selectively correlate (another useful property!).

One my concerns with selective disclosure in SD-JWT/CWT/ecdsa-sd/mDoc is
that only the issuer can choose what fields are selectively disclosable,
which IMHO they will only do if it is in their interest, which may not
necessary be the interest of the subject, or other holders (for instance,
an employer holding a employee subject credentials may have other needs to
elide that are different than the issuer and the subject). One particular
consequence of this is that there may be very few fields in a credential
that are selectively disclosable. Combined with various approaches for
"holder binding", things become even more challenging. There are also some
questions about when it is appropriate to do BBS to also anti-correlate
signatures — there are cases where it might not make sense. No easy
solutions!

-- Christopher Allen