Re: Changes we'd make to EPP, was Re: Domain check in draft-obispo-epp-idn-00.txt

2012/1/17 Patrik Fältström <[email protected]>:
> On 17 jan 2012, at 09:48, Theo Kramer <[email protected]> wrote:
>
>> I can see none. Garbage collection is a function of registry maintenance imo.
>
> A registry can not know if an existing contact record is to be connected to a domain name that is to be created, as by definition contact records must exist with the registry that is not connected to any domain object.

Indeed.

When I have brought this up, the focus is always on the business
policy side, not the technical policy.  Everyone gets very concerned
about the notion that we might be destroying data that "belongs" to
the registrars.

And yes, indeed, we don't know if a record is about to be used.
(Though if it has not been connected to anything for several years,
one might take a pretty strong hint that it has been discarded...)

My favourite idea, which has seen zero up-take, has been the notion of
creating a policy along the lines of "Registrars are allowed to have a
certain number of unused contacts 'for free.'  If the number exceeds
that threshold, we'll bill them something, that'll encourage them to
clean up their unused contacts."  The bit of economist in me likes
that form of feedback, and I certainly don't have in mind for this to
be a profit centre, just to make sure that registrars find it of value
to clean up after themselves.  Pushing that through the various layers
of politics seems pretty daunting.

Alternatively, it would be in principle plausible to add some logic
surrounding the LINKED status, to put a TTL onto contacts any time
they change to be unlinked.  What should initiate GC, and how to
communicate the work done back to the registrars is a good question.

Your idea of a "GC" command is certainly interesting, though I see
things to have reservations about:

a) Why should a registrar want to issue this command in the first
place?  I suspect we have quite a few registrars that never care to
pull data coming from their poll queues, and would expect some overlap
between those that don't care about that and those that would never
bother issuing a <epp:garbage-collect> command.

b) What should the return result be?

  1.  Perhaps merely a "Yep, we'll be getting on to that" response.

  2.  If there are 500000 contacts to trim out, I don't imagine it's
reasonable to expect for this request to wait until that is all done.

  3.  It might be nice to get a list of trimmed contacts, but if there
are 500000 of them, that's not likely reasonable.  Proposals for EPP
extensions offering very large response sets have generally gotten
push-back.

  4.  Perhaps the list of contacts trimmed get thrown into the poll
queue.  But 500000 of them mayn't be reasonable to process.  That may
just shove the problem around a little.  Fixing the "dead contacts
problem" may lead us into having much worse problems with poll queue.
_______________________________________________
provreg mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/provreg