> To: The structure of the RPKI is modeled on the existing organizational > structure that is already responsible for IP address and AS number > resource allocation. In this allocation hierarchy, IANA allocates > resources to five Regional Internet Registries (RIRs), each of > which
the iana has, and will continue to have, direct allocations to others than the rirs > To: The IESG MUST approve a replacement BCP that either updates or > obsoletes this BCP, following the procedures of the IETF Standards > Process as defined in RFC 2026 [RFC2026]. the ietf must approve, by whatever means it currently uses as an approval process. do not forget 1992, and we seem to have bread massively more process wonks. > Provider-independent (PI) address space - PI address space is > allocated directly by an Internet Registry to a subscriber. This space > cannot be further allocated by the subscriber to other entities. i do not believe that last condition is firm, especially for legacy space. as a well-known example, 135.whatever was split when bell split. > The subscriber must contract with an ISP to obtain routing of the > address block within the Internet. the holder of pi space may very well be an isp > 2.2 Publication of certification information > > Added: > Each CA MUST publish its CPS. and up above, you say a ca is someone who makes an ee to sign a roa. most of them just ain't gonna publish a cps. and when an rir or large isp acts as a proxy privatekey/cert holder for a customer and generates a roa for them (the current rir model), who specifically is the ca that will publish their policy? > To: The Subject name in each certificate does not need to be meaningful. > The certificates issued under this PKI are used for authorization in > support of applications that make use of attestations of Internet > resource holdings, not for identification. The PKI binds a public > key to sets of INRs. The name of the holder of the INR is not > intended to be "meaningful". i am confused on the status of this. once upon a time it was must not be meainingful. then there was a move to make it meaningful for rirs but no one else. and now? can uu issue child certs with customer subnect names? should they? > However, Subject names in certificates SHOULD be constructed in a > way that minimizes the chances that two entities in the RPKI will be > assigned the same name. and how they heck will that be done? and why? > To: Since Subject names are not intended to be meaningful, there is no > provision to recognize or authenticate trademarks, service marks, > etc. When generating subject names, CAs should use a mechanism that > is unlikely to result in Subject names that include trademarks, > service marks, etc. CAs MUST NOT knowingly issue a certificate with > a Subject name that a court of competent jurisdiction has determined > infringes on the trademark of another individual or organization. great! and my sn generator should check the 200+ trademark databases of the world to be sure blorzlpktz is not trademarked? like hell. how about just the stuff from the beginning, do try not to generate meaningful names, period. taking a position on enforcing trademarks is something we learned to stay many kms away from dns registries. > Added: This section describes the procedures for certificate re-key. > Certificate re-key is the issuance of a new certificate to replace an > old one because the key needs to be replaced. Unlike with certificate > renewal, the public key is changed. s/with// > To: This PKI does not make provision for use of OCSP or SCVP, because it > is anticipated that the primary RPs (ISPs) will acquire and validate > certificates for all participating resource holders. These protocols > are not designed for such large-scale, bulk certificate status > checking. RPs MUST check for new CRLs at least daily. It is > RECOMMENDED that RPs perform this check several times per day, but > no more than 8-12 times per day (to avoid excessive repository > accesses). i am not sure i am comfortable with the timing here. first, does this cover where a local relying cache pulls and validates from a 'parent' trusted cache (note these are not publication points)? secondly, this places a lower bound on repairing a hijack or mistake. while, admittedly, the 12 times a day might not be an even distribution, it is sure to be interpreted that way. but i am not sure i see a fix. > 5.1. Physical controls come on. this is a bit of a pita for widget corp who has a cert for their space and wants to issue a roa. and i am not sure a ca with security or privacy concerns is gonna want to say anything. > 5.2. Procedural controls > Added: Topics covered SHOULD include: same issues as 5.1 for folk with security or privacy concerns. > 5.4.1. Types of events recorded -- 2nd sentence > . Clock adjustments got a logging ntpd? > To: In the RPKI, each subscriber acts as a CA authoritative for the > specified INRs that were distributed to that entity. which answers my question above. if i am a holder with a parent (isp or ror) acting as my key/cert proxy, then i am still a ca. i do not disagree with this. but note that puts all the cp obligations on jillions of end holders. randy _______________________________________________ sidr mailing list [email protected] https://www.ietf.org/mailman/listinfo/sidr
