Minutes of the interim virtual meeting of the IETF DNS Extensions
Working Group

2010-02-16 15:00 UTC

via teleconference and Webex session.  Approximately 35 attendees.

The meeting was chaired by WG co-chair Andrew Sullivan.

The chair opened the meeting, apologised for some unexpected
technical limitations.  The chair reminded everyone that the meeting
is an IETF activity, covered by the Note Well statement and the
relevant IPR rules, and that participants should "sign the blue
sheets" by sending email to the chair noting their participation.
The chair took names from the Webex session, but may not have access
to the email addresses associated.

The chair outlined the goals of the meeting, and opened the floor for
discussion.  Remarks in favour of there being a problem to solve came
from Vaggelis Segredakis, Paul Hoffman, and Sotiris Panaretou.

The chair asked participants whether anyone thought there was
not a problem to solve.  The chair heard nobody argue for this.

The chair asked about acceptable limitations.  Does anyone think
that what is really needed is a replacement of the DNS protocols?
There were no responses.

The chair next asked whether it is acceptable that zone administrators
need to understand whether they are in a "canonical chain" or not.
There was some discussion around this matter, with the initial
observation that it was how things had to work today anyway.  Further
discussion illuminated that the problem persists down the chain,
because "variant" children down the tree might not realise that they
are working in a variant space, and that there may be problems as a
result.  Jim Reid worried that this could lead to a "hall of mirrors"
and that the best answers therefore all lie in provisioning rather
than DNS responses.  This led naturally to the presentation of the
zone clone idea from Paul Vixie.

The presentation offered a number of questions of detail that would
need to be settled in order to standardize the zone clone proposal.
Paul Vixie included some constraints on solutions in his presentation:

-Must not require stubs or recursives to be upgraded, since there are
millions of these and the tail is long

- Must be an Internet Standard, not a proprietary or adhoc extension,
to facilitate multivendor operation

- Must not place any burden on registry, which may be regulated (so,
autoinsertion into root zone, no!)
 
- Authority server operators, protocol implementors, and registrars
can accept burdens, since they have incentives, and are few in number

After discussion, the chair asked for a sense of the WG.  There
was considerable support heard for the constraints, and no
opposition.  These therefore appear to be useful constraints on the
WG's plans.  The chair understood the WG's desire to apply generally
to work in this area, but subsequent conversations clarified that many
participants thought the constraints to apply just to zone clones.  

The chair summarized what he had heard from the WG during the meeting.
It appears that the WG has eliminated "do nothing", "provisioning
only", "CNAME-only" or "DNAME-only" approaches.  There was little
discussion of the BNAME proposal.  There was no discussion of altering
DNAME to allow inclusion of CNAME in the answer.  There was no
discussion of other approaches.  So, from the WG's Wiki outline, items
1, 2, 4, 5, and 6 all seem to be ruled out.  Item 3 appears to be an
option.

Items 7-9 are inconsistent with Paul Vixie's set of conditions.  The
WG seemed to indicate that those conditions were the right ones.  At
the meeting, the chair noted that items 7-9 had not been discussed,
but since they are inconsistent with the apprehended view of the WG,
they're not live options.  Everyone may not have realised this
consequence [note: the chair didn't at the time] , so the WG chairs
will treat this entailment as unsettled for the time being.

The chair took names of people who are willing to work on the problem.
There was enough indication of support to suggest that the WG might
want to tackle the problem.  The next steps are to get solid drafts to
evaluate, to ensure the problem statement is correctly narrowed in
time for Anaheim, and to advertise part of the Anaheim meeting widely
to attract as many interested parties as possible, in order to check
assumptions before the WG progresses too far on this work item.  If
that meeting is successful and clear in its outcome, then this item
will be adopted as a WG work item with clear milestones.