Hi, David Cooper wrote: "Contrary to what the defect report says, X.509 intentionally defined CRL to include revocation lists that covered public-key certificates and revocation lists that covered attribute certificates," I been holding DR_391 back for that reason. What comes from NIST, I take seriously. I have now updated the DR to a more extensive discussion without suggesting a resolution. Please comment on which way to go, as the answer is not that obvious (see http://x500standard.com/uploads/Ig/DR_391.pdf). I still need to address the concern about the issuingDistributionPoint extension as raised by David. Erik From: x500standard-bounce@xxxxxxxxxxxxx [mailto:x500standard-bounce@xxxxxxxxxxxxx] On Behalf Of David A. Cooper Sent: Tuesday, January 28, 2014 3:48 PM To: x500standard@xxxxxxxxxxxxx Subject: [x500standard] Re: Revocation list definitions On 01/28/2014 07:19 AM, Erik Andersen wrote: Hi, To separate PKI and PMI within X.509, I have made a proposal for updated certificate revocation definitions. Please comment! Erik Erik, I am against this defect report as well. However, I've noticed a far more serious problem. I see absolutely no reason to insist on a complete separation between PKI and PMI, but it appears that non-backward compatible changes have already been made to move in this direction. Contrary to what the defect report says, X.509 intentionally defined CRL to include revocation lists that covered public-key certificates and revocation lists that covered attribute certificates, so it cannot be "clarified" to make it clear that CRLs only cover public-key certificates. Even now there is still text in X.509 that says otherwise: The CRL distribution points extension shall be used only as a certificate extension and may be used in authority-certificates, end-entity public-key certificates, and in attribute certificates. This field identifies the CRL distribution point or points to which a certificate user should refer to ascertain if the certificate has been revoked. A certificate user can obtain a CRL from an applicable distribution point or it may be able to obtain a current complete CRL from the authority directory entry. Note that the term "CRL" is used even though the extension may appear in either a public-key or attribute certificate. What I find very disturbing, however, is that I just noticed that the definition of the issuingDistributionPoint extension has been changed (without changing the OID). In the 4th edition it was defined as: IssuingDistPointSyntax ::= SEQUENCE { distributionPoint [0] DistributionPointName OPTIONAL, onlyContainsUserCerts [1] BOOLEAN DEFAULT FALSE, onlyContainsAuthorityCerts [2] BOOLEAN DEFAULT FALSE, onlySomeReasons [3] ReasonFlags OPTIONAL, indirectCRL [4] BOOLEAN DEFAULT FALSE, onlyContainsAttributeCerts [5] BOOLEAN DEFAULT FALSE } In the 6th edition it is defined as: IssuingDistPointSyntax ::= SEQUENCE { -- If onlyContainsUserPublicKeyCerts and onlyContainsCACerts are both FALSE, -- the CRL covers both certificate types distributionPoint [0] DistributionPointName OPTIONAL, onlyContainsUserPublicKeyCerts [1] BOOLEAN DEFAULT FALSE, onlyContainsCACerts [2] BOOLEAN DEFAULT FALSE, onlySomeReasons [3] ReasonFlags OPTIONAL, indirectCRL [4] BOOLEAN DEFAULT FALSE } RFC 5280 contains the definition from the 4th edition, and I don't see the IETF updating RFC 5280 to change this. I don't know what (if any) other non-backward compatible changes were made in X.509 as part of an effort to "separate PKI and PMI within X.509," but making unnecessary non-backward compatible changes like this to a stable standard can do nothing but cause problems. Dave ----- www.x500standard.com <http://www.x500standard.com> : The central source for information on the X.500 Directory Standard.