Re: Questions on Authority/Subject Key Identifiers

["Tom Gindin" <tgindin@xxxxxxxxxx>]

      Peter:

      GIven the arbitrary nature of KID's, why is SKID/AKID chaining in the
absence of DN matches even a sensible thing to do?  If everybody used
method 1 in RFC 3280 section 4.2.1.2 (or maybe even method 2, since 2**60
is far beyond the square of the number of CA's in the world), it would be.
However, the comment about "unique values" doesn't seem to have enough
teeth to avoid collisions, and the "monotonically increasing sequence of
integers" technique which is explicitly considered reasonable by RFC's 2459
and 3280 makes collisions between different issuers fairly likely.  X.509
just says they have to be unique in an issuer's space.

            Tom Gindin

pgut001@xxxxxxxxxxxxxxxxx (Peter Gutmann)@mail.imc.org on 09/29/2002
01:52:22 AM

Sent by:    owner-ietf-pkix@xxxxxxxxxxxx


To:    kent@xxxxxxx, pgut001@xxxxxxxxxxxxxxxxx
cc:    ietf-pkix@xxxxxxx
Subject:    Re: Questions on Authority/Subject Key Identifiers



Stephen Kent <kent@xxxxxxx> writes:

>The real world issue you describe calls for re-issuance of certs; it does
not
>justify violating the standards.

In *theory* it calls for the re-issuance of certs.  I can quite easily see
that the practical approach would be to chain by sKID, and obviously enough
people agree with this that they persuaded MS to change the behaviour of
their
code to allow it (the issue you're referring to was a bug, not a deliberate
design decision like sKID chaining).

Peter.