[dmarc-ietf] Rollback

[Douglas Foster <[e-mail]>]

Replacing the PSL was an interesting line of investigation, but it was not
part of the charter, so it can only be part of DMARCbis if it works well,
which it does not.

For non-PSL to work, all private registries (which are currently and
correctly listed in the PSL) must tag themselves in the DNS.
- How long will this take?   A long time, maybe forever.
- If it happens, will evaluators know that they can trust that it happened?
No, never.

At best, we have an idea for an experimental algorithm, not a
proposed standard, and our group is so small that we lack a quorum of
willing participants in the experiment.   The PSL replacement idea, and all
of the supporting language, needs to be abandoned.

It is time to roll back.  Instead of discarding the PSL, we need to fix its
weaknesses.   The PSL may contain errors, and RFC7489 does not provide a
way for DNS administrators to document and correct those errors (or confirm
correct results.)    We could define that error correction mechanism.

Possible PSL results:

Lands too high errors:
PSL+1 is another PSL, not the organization domain
PSL+1 skips over a private registration, and returns the registrar domain
instead of the client domain.
PSL+1 skips over a private registration, and returns a subdomain of the
registrar domain instead of the client domain.

Lands too low errors:
PSL+1 returns a subdomain of the organizational domain

Lands just right but is still an error:
PSL+1 is non-existent because it is not registered with the parent domain.

Correct results:
PSL+1 is an organization domain of the correct organization, and is
confirmed with a DNS indicator
PSL+1 is not explicitly confirmed or explicitly rejected, so it is presumed
to be the correct organizational domain.

Doug  Foster