A Publication Protocol for the Resource Public Key Infrastructure (RPKI)
Summary: Has 3 DISCUSSes. Needs 4 more YES or NO OBJECTION positions to pass.
Alissa Cooper Discuss
What is the upgrade path for the future when new versions of this protocol get published? How are clients and servers meant to agree on which version to use?
(Hit the send button too quickly, sorry for the multiple emails.) Although I understand why Section 6 says transport security is not strictly required, given that the authentication and authorization mechanisms that this protocol relies on are outside of the scope here, isn't it possible that clients and servers may be exchanging cookies or other headers in the course of using this protocol that would benefit from transport encryption? It seems like mentioning that transport security may still be beneficial although not required might be a good idea.
Terry Manderson Discuss
Updating after reading sidr-oob-setup-06. I see that the publisher wins as per Section 5.2.4 <repository_response/> My original discuss us below, so you may omit the concern about negotiating the URI. The remainder stands. Thanks T. =-=-=-=-=-=-=-=- Thanks for finally bringing this protocol forward. I support Alissa's and Alexey's concerns. I only have one discuss for this draft. Looking at section 4, operational considerations I was expecting to see a review of any considerations as to how this protocol works, the interaction between the layers of HTTP, CMS, and XML and any implementation differences/difficulties that exist between the 2 known implementations. Instead there is a discussion on laying out the repository structure under the mandatory to implement _retrieval_ mechanism (RSYNC) and the nuances of RSYNC itself. This appears to be misplaced as the protocol (HTTP/CMS/XML) interactions here are simply about publication from a certificate authority operator to a repository operator, and in that space surely the publication protocol (this doc) is agnostic to the exact repo structure. In both a database world (not a file based one) and where multiple RPKI fetch mechanisms (rsync, http, torrent, etc ...) are used, how is the exact URI meaningful for sidr-publication? There might be a deeper problem here regarding any potential collisions and negotiation of the URI space between the certificate authority operator and the publication repository operator. (sure, in a situation where Alice does both, no problem.) So you may wish to address issues like that in the operational considerations section as opposed to dealing with RSYNC (in)efficiencies.
Alexey Melnikov Discuss
I support Alissa's DISCUSS point on versioning. I find the document to be a bit short on normative references and some implementation details. Other than that the document looks fine. My specific questions and concern are as follows: 1) Please add a normative reference for HTTP, URI and RelaxNG on first use. 2) Base64 needs a normative reference (including the section number, as there are 2 variants). 3) Section 2 says that all payloads use CMS. None of your examples show CMS. Can you please elaborate on how CMS is used? 4) How can URI of the service be discovered?
In 2.5: is the list of error reasons extensible? Was Relax NG schema validated with a tool? In Section 5 you should reference this document (and not just section numbers), as IANA registrations cut & pasted to IANA website as separate files.
Alvaro Retana Yes
Alia Atlas No Objection
Deborah Brungard No Objection
Spencer Dawkins No Objection
Mirja Kühlewind No Objection
My only question is why this is a sidr wg doc? This seems like a general mechanism that cannot only be used in the routing infrastructure. Has this doc been at least reviewed by other wgs?
Kathleen Moriarty No Objection
Thanks for addressing the security directorate review. As for Alissa's comment on transport, more language added to the Security Considerations section would be helpful to explain why the CMS signature is sufficient. I am assuming that the only exposure would be to public information during transport that is protected from tampering, unless I missed something in reading the draft (I don't think you are transferring private keys and didn't see that in the text). Security controls being managed according to the CA policy mentioned earlier in the document is appropriate, having run CAs before - there are strict requirements already depending on the level you plan to run the CA.