Re: [keyassure] Interpreting certificates (and summary)

[Zack Weinberg <[e-mail]>]

On Wed, Feb 23, 2011 at 2:01 PM, Paul Hoffman <[e-mail]> wrote:
> On 2/23/11 1:03 PM, Zack Weinberg wrote:
>>
>> Could we short-circuit this entire argument by deferring to the
>> protocol being secured for the certificate format?
>
> The WG already agreed that "the protocol being secured" is TLS, as the
> current draft states. TLS has exactly one certificate type: PKIX
> certificates in the format given in RFC 5280. (There is a
> non-standards-track extension to allow OpenPGP certificates.)

Right.  My suggested edit isn't meant to be a semantic change, just to
arrange that we can stop arguing about what that certificate format is
called.

... It does also mean that OpenPGP-cert extension, if it ever becomes
popular, could also be accomodated by cert types 1 and 2, because they
would still be "certificates in the wire format used by the protocol".
 Personally I see that as a good thing, because (one fervently hopes)
the OpenPGP format is unambiguously distinguishable from the X.509
format, so putting that information into the RR metadata would be a
redundancy and open the possibility of the two being out of sync.

... Come to think of it, the distinction between cert types 1 and 2
(end entity or CA) is *also* a redundancy (with the isCA field and/or
the key usage restrictions) that should perhaps be eliminated.  Why
not just use type 1 for both, and have the client's processing pay
attention to the cert's self-description?  (This is not a rhetorical
question; maybe there's a good reason I don't know about.)

zw