On Thu 17/Mar/2022 20:50:39 +0100 John Levine wrote:
It appears that Alessandro Vesely  <[email protected]> said:
To find the org domain for a domain:
   chop the domain to the last five labels and walk up the tree.
   stop when you find a DMARC record with psd or you hit the root.
   if a record has psd=n, that's the org domain
   if a record has psd=y and it isn't the original domain, the org domain is 
the one below it
   otherwise the org domain is the last (highest) DMARC record you found

Fine.  DMARC /specifies/ that the org domain MUST publish a record.

No, that's not what it says.  If you encounter psd=y and go back one,
that can be an org domain without a record.  It is currently possible
to have org domains without records and I see no reason to change that.


The reason to require an org record is that filters cannot determine that a domain is an org domain without it. The only exception is domains registered right below a PSD publishing psd=y. Otherwise the org domain is the last (highest) DMARC record you found.


Relaxed alignment doesn't change, if two domains have the same org domain, 
they're aligned.

On a mail From:[email protected], assume we have already determined that the org domain is c.d. Then there is a signature with d=e.f.c.d. It is aligned based on string comparison.

Repeating the tree walk, we'd get a different result if we find psd=y at _dmarc.f.c.d. Is that realistic?

Yes, of course it is.  Look at some of the PSL entries.


I may be dumb, but I stared at the PSL for a few minutes and couldn't think of non-aligned subdomains of an org domain. Yes, there are PSDs which are a substring of another PSD. For example .it and .milano.it, or *.kobe.jp and !city.kobe.jp. However, prepending the org domain label you obviously get a unique suffix. Everything at or below the org domain belongs to that org. Any two identifiers belonging to the same org are mutually aligned.


Best
Ale
--






_______________________________________________
dmarc mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dmarc

Reply via email to