Steve,

Stephen Kent wrote on 15-10-2009 21:28:
> Folks,
> 
> At the RIPE meeting in Lisbon last week, there was some discussion about
> how relying parties (e.g. ISPs) might like to see reason codes used in
> the CRLs issued in the RPKI. The concern motivating this is a situation
> in which a CA (e.g., an RIR or an ISP) might be forced to revoke a cert
> for a customer because of some form of legal pressure, e.g., an effort
> to shut down a phishing site. The argument is that if an RIR were able
> to insert a reason code in the CRL entry for such a revocation, an ISP
> would be able to decide whether to accept or ignore the CRL entry in
> question. The relevant X,509 and PKIX standards allow for such a
> capability, but we removed it for the RPKI cert/CRL profile, because we
> didn't see a need for such codes.  The codes can be used on a per-CRL
> entry basis, so only when a cert is revoked for a reason that the AC
> elects to note would a code be inserted into the CRL entry for the
> revoked cert.
> 

I thing such feature may be useful as a general communication capability
of a CA to the RPs. Although maybe not exactly as a secret protocol
behind the LEAs backs :)

However, I don't think we should go ahead and modify the profile, mainly
for the reason that we won't be able to get it right at this point in
time. I don't think we can create realistic scenarios for usage of this
feature, and consistent interpretation of the reason codes is essential
(as opposed to per CA/CPS).

If we feel that we may not have enough time to introduce this feature
when needed, I am fine with reserving this in the profile, but providing
no guidance on how to process/interpret it (or rather advise not to
process the code).

Out of curiosity - how do the reason codes inform the validation process
in the conventional PKI?

> 
> (As a side note, I made a brief presentation describing how cert
> validation software for the RPKI could offer RPs various options for
> dealing with expired and revoked certs, to address similar concerns.
> These options do not require any changes to the cert/CRL profile, but
> they also do not offer an automated capability for selective CRL entry
> processing.)
> 
> There are some costs associated with offering this option:
> 
>     - we make CRL processing more complex for all relying parties
> (relative to the current profile). the reason is that if we include this
> feature as part of the profile, ALL RPs MUST be prepared to process CRLs
> that make use of the feature. There is no requirement that ANY CA make
> use of the feature; it would be up to each CA and would be reflected in
> the CPS for the CA
> 
>     - it is not clear if law enforcement authorities would allow a CA
> (e.g., an RIR or an ISP) to mark a CRL in this fashion once they
> understood the potential implication. The CA could cite its CPS as
> mandating that it follow this procedure, as a defense of the practice,
> but who knows if this would suffice.
> 
>     - reason codes are defined by an ENUMERATED data type.  This means
> that there is a list of codes, each identified y an integer, and no
> provision for defining private reason codes. The list was motivated by
> the financial services industry, so there i no code that matches the
> notion of "the men in black made me do it." I note that the value "7"
> has no reason code assigned to it, but to use it for this purpose would
> be wrong :-).
> 
> So, before we ask Geoff to modify the Cert/CRL profile to allow for
> reasons codes, I'd like to get some sense of whether  the WG feels this
> is a reasonable path to follow.
> 
> Steve
> _______________________________________________
> sidr mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/sidr

Andrei
_______________________________________________
sidr mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/sidr

Reply via email to