Last Call Review of draft-ietf-dime-rfc3588bis-
review-ietf-dime-rfc3588bis-secdir-lc-wallace-2010-11-30-00

I have reviewed this document as part of the security directorate's ongoing
effort to review all IETF documents being processed by the IESG.  These
comments were written primarily for the benefit of the security area
directors.  Document editors and WG chairs should treat these comments just
like any other last call comments.



This is a fairly long document and my review has been conducted in chunks,
answers to some of the questions may be apparent on a subsequent reading.  My
primary concern is with authorization of nodes.  Comments, questions and a few
nits are below:



- Section 2.1 should state TLS usage requires mutual authentication.  At
present, this is not mentioned until section 13.



- In 2.6, there appears to be a stray sentence fragment: "Additional security
information, when needed (e.g., keys, certificates)".  Suggest adding a
sentence two beneath if this is another item in the list.



- In section 2.9, authorization checks are mandated but no mechanisms are
defined.  How is authorization performed?  Is this related to the authorization
checks described in 2.9?



- What does this mean: "the domain name in the SRV query and the domain name in
the target in the SRV record MUST both be valid based on the same site
certificate"?



- Should expand CA acronym in 5.2.



- In 5.2, including OIDs in a certificate is not a useful authorization
mechanism unless those values are processed across the path from the trust
anchor to the server certificate (or there are some restrictions on the nature
of the path or trust anchor store).  What OIDs are meant here, extended key
usage or certificate policy OIDs?



- In 13.2, Are trust anchor stores per realm?  If not how will relays be able
to determine how to use authorization information?



- May want to reference RFC 5280 to ensure full path validation (including
revocation status determination) is performed.



- In 5.1, the first bullet correlates to the state table in section 5.6.  The
second bullet does not.  Should it?



- Is the deprecation in 6.10 worth a SHOULD NOT?



- The security considerations should include some words describing the
consequences of inappropriately authenticating or authorizing a diameter node.



- Should relays, agents etc. be required to authenticate to clients with the
same identity used to authenticate to servers?