On Fri, Jan 23, 2015 at 1:21 PM, Stephen Farrell <[email protected]>
wrote:

>
>
> On 23/01/15 18:12, Richard Barnes wrote:
> > Yes, you could quasi-templatize it.  You would still have to manually
> > adjust length fields and munge byte strings together.  Eww.
>
> Not in practice no, public key lengths aren't that variable
> and all that's is done by most libraries incl. webcrypto and
> openssl - in fact I'd be interested in knowing which crypto
> libraries do not directly support this already. I can't think
> of any tbh. And I assume our thumbprinting code will use
> a library for the hashing and that the public key is very
> likely to be used as well as just thumbprinted, so your
> argument seems to me broken and h(spki) likely simpler
> overall.
>
> > Given the amount of DANE deployment, it's not a huge priority for me.
> DANE
> > could also meet us half-way here by defining a new binding type should
> > there be a need.
>
> It's not only DANE - if you go looking around h(spki) is
> what's used elsewhere too. And the reason is it's obvious
> and correct.
>
> The choice seems to be between something compatible with
> other protocols (allowing comparisons) or rolling your
> own to save 2 lines of code or so. (Assuming there are
> no other advantages to what you're suggesting?)
>
> But we're repeating ourselves nearly I guess.
>

As a matter of process, this would have been a good objection to bring up
at milestone / doc adoption time, since if we're just doing h(spki),
there's no need for a document.

--Richard



>
> S.
>
>
_______________________________________________
jose mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/jose

Reply via email to