[DNSOP] ANAME precedence [issue #58]

[Matthijs Mekking <[e-mail]>]

Hi,

I would like to start separate threads on the remaining issues of the
ANAME draft. One issue that remains to be solved is whether having an A
or AAAA record next to the ANAME should take precedence or not.

  Draft: https://datatracker.ietf.org/doc/draft-ietf-dnsop-aname/
  Issue: https://github.com/each/draft-aname/issues/58

This was discussed face to face during IETF 101 and at that time the
conclusion was that the correct behavior is that ANAME takes precedence:
If you implement ANAME, the target lookup for A and AAAA will always be
made. If the lookup succeeds, the sibling address records are replaced
with the target address records. If the lookup fails, the sibling
address records remain in the zone.

Jan Včelák mentioned that at least NS1 uses a different order of
priority: If an sibling address record exists next to the ANAME it takes
precedence and no target lookup is done for that address record type.

In order to provide identical behavior between providers (make ANAME
work in the multi-provider model) we should agree on the priority order.

To me, it makes much more sense to use the sibling address record as a
default, and the ANAME target lookup can replace the sibling address
records. The target address records will improve the answer.

If you place an override, adding an address record next to ANAME, you
can achieve the same thing by not placing the ANAME record in your zone
at all.

But when the sibling address records take precedence, it has the
property that you can set up ANAME for only one address type, for
example ANAME for A but not for AAAA. I would like to know if there is a
good use case for having this property.

I would like to hear an opinion from the working group (preferably from
ANAME providers). Specifically do you have a preference of priority
order? Do you think having the "set up ANAME for one address type"
property is worth having?


Thanks,

Matthijs