Domain Registration Protocol BOF @ IETF-43
Dave Crocker / DC
Kent Crispin / KC
Eric Brunner / EB <email@example.com>
Dave's opening remarks discussed the time-line for effective work on the issue of shared registries (aggressive), the need to form a WG charter in the 2 hours allotted for the BOF, and to select a forward-progress goal. These issues were re-visited towards the end of the BOF, motivated by the observations of the two ADs present, and the sense of the BOF attendees.
Dave gave an intro to the issue(s) and the draft, running through the first six (6) slides. #2 introduced the SRS (shared registry service) terminology, #3 introduced the two basic models -- light vs heavy, and slides #4 and #5 discussed the salient details of these two models, with slide #6 covering the trade-offs of each.
Kent's first slide and oral history covered the issue since new-dom. The second slide covered the general architecture of the SRSP, and elicited a comment concerning the need for ccTLD-specific extension, specifically the insertion of a "Registry Authority" in the SRSP sequence of transactions. The authority would mediate the transaction.
This was characterized (EB) as a non-delegated transaction to an external authority, similar in nature to any name conflict resolution transaction, and also characterized (DC) as a policy issue for which the SRS protocol needed only to provide a mechanism.
Sustained discussion on several basic issues occurred at this point, in summary the issues of ccTLD-specific- vs gTLD-specific policies, equity of access within a registry, and name portability within a registry took place. This clarified the distinctions between registry and registrars, and the assumption that gTLD requirements could be extended to meet those of ccTLDs.
The outcome of this phase of the discussion was that
1. if necessary to distinguish, the WG scope is gTLD specific,
2. many registrars acting on hierarchies of gTLD-rooted registries
3. equity of access within each registry is desired
4. name portability "within" each registry is desired
Kent continued with slide #3 covering registry data and slide #4 covering access functions. There was some terminology haggling and "inquire" was substituted for "access" function. A question was posed on the internal representation of registries and the "format free" burden one registrar could place upon subsequent registrars upon change-of-registrar by users.
A question was posed on the authentication mechanism(s) that ran into the content of Kent's slide #5 (security), arising from the outcome of separation of registries from registrars. This question was followed up by several more related to the existing SRS authentication issues. One of the two ADs attending (Patrik) observed that we should not require the use of a risk-prone authentication mechanism.
Kent continued with slide #6 (transport), a summary of questions and answers follows:
a) is there a performance requirement for the protocol
1. scale (registry deltas) to 15 mins granularity?
ans: (DC) scale to arbitrary numbers, hence yes generally.
ans: (KC) key verification is the performance bottleneck, hence not "dynamic DNS".
2. geographic and other extensions?
ans: (DC/KC) additional objects and fields suffice, hence no.
c) no coupling of protocol and transport (design principle assertion from the floor)
ans: (DC) email as transport is a conscious choice to mandate a minimal level of services required, consistent with basic principle of equity of access, and "SHALL" is appropriate.
d) another ccTLD comment, which EB understood to again seek some form of two-phase commit (external authority hook)
e) is the SRS model wanted below the level of 2SLDs?
ans: (DC/KC) yes, recursive generally.
At this point AD Patrik made a comment with respect to priorities and the ccTLD vs gTLD choice for the WG, in brief strongly advising that the gTLD issues are more pressing.
Only 10 of the approximate 50 attendees had seen the draft charter, so the draft charter was read. There was more discussion of the ccTLD vs gTLD issue, of ASCII vs UTF8, and email as "a" (not "the exclusive") mandated transport.
KC observed that the issue of ASCII vs UTF8 had been discussed previously on the mailing list, and that the question was still open.
The issue of registrar-specific variable/private data within registries again was discussed (see "format free" above), as was the external authority issue, and DC suggested that "extensibility", along with specific wording changes to the charter be taken to the mailing list.
It was the sense of the room that the mandate for email as transport was sound.
Dave discussed the motivation for moving fast on the issue, followed by point by one of the ADs that this WG cannot form until January for IESG-internal reasons.
It was the sense of the room that the existing draft and charter, with the changes arising from the BOF, were sufficient to forward on to the ADs for action, and that the choice to act "fast" was adopted.
Outstanding Issues (KC)
1. Extensibility (but not dynamic DNS), KC "easy to add"
2. Policy (firewalls and external authority)
3. Dynamic private, KC "hard"
The BOF ended with 20-30 people committing to work on the documents and 10+ people committing to "really working" on the texts over the immediate future to meet the "fast" time-line requirements.