Sean,

Thanks for the quick review. Replies inline...

On 12/5/12 4:30 PM, "Sean Turner" <[email protected]> wrote:

>Andy,
>
>A couple of comments:
>
>1) I'm hoping to constrain the type and number of qualifiers that can be
>included.
>
>5280 defines two types: cps (for certificate practice statements) and
>unotice (to display info to relying parties when the certificate is
>used).  I'm hoping you just want the cps choice, which is just a URI.
>And, if it's just the CPS then there's only one CPS under which a
>certificate is issued - right?  How about:
>
>OLD:
>
>  This document updates [RFC6487], Section 4.8.9, to explicitly allow
>  optional PolicyQualifierInfo objects in the PolicyInformation
>  specified by [RFC6487].
>
>NEW:
>
>  This document updates [RFC6487], Section 4.8.9, as follows:
>
>  OLD:
>
>    This extension MUST be present and MUST be marked critical.  It
>    MUST include exactly one policy, as specified in the RPKI CP
>    [RFC6484].
>
>   NEW:
>
>    This extension MUST be present and MUST be marked critical.  It
>    MUST include exactly one policy, as specified in the RPKI CP
>    [RFC6484].  Exactly one policy qualifier MAY be included.  If a
>    policy qualifier is included, the policyQualifierId MUST be the
>    CPS pointer qualifier type (id-qt-cps).
>
>I think it's clear the value is the cPSuri choice, do you think anybody
>else would pick userNotice?

It is possible that somebody somewhere might find them useful. But I'm not
gonna fall on my sword over the inclusion of user notices. A CPS pointer
is what we need.

I'll incorporate your text. Thanks.

>
>3) Two process points:
>
>3.1) Need an IANA considerations section:
>
>IANA Considerations
>
>None.

Noted.

>
>3.2) Need a security considerations section.  It would also be good to
>say why it's not a security issue to add the URI, but you'll need to
>confirm my assumption that relying parties aren't actually going to
>chase the URI.  Alternatively, text could be added to s7.1.1 of RFC 6487
>to say "don't process the URI", but I think putting it in the security
>considerations is probably less painful.  Suggested text:
>
>Security Considerations
>
>The Security Considerations of [RFC6487] apply to this document.
>
>This document updates the RPKI certificate profile to specify that the
>certificate policies extension can include a policy qualifier, which is
>a URI.  Checking of the URI might allow denial-of-service (DoS) attacks,
>where the target host may be subjected to bogus work resolving the URI.
>  However, this specification, like [RFC5280], places no processing
>requirements on the URI included in the qualifier.

This is a very good point. And I think addressing it in security
considerations, as you have suggested, is the appropriate thing to do.

>
>4) I hope you'll ask the WG to adopt this draft ;)

Yes, I was planning to do so after a re-spin of this document.

Thanks for your review and text.

-andy

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

Reply via email to