It appears that Murray S. Kucherawy  <[email protected]> said:
>The tree walk might be the DBOUND solution, for all we know.  Having it in
>a separate, generic-as-possible, document might make the technique usable
>by other applications as well.

We had a few plausible proposals in the DBOUND group, and none of them
were anything like a tree walk. Most of the things that use the PSL
are related to browser UI and do not have their own DNS records. I
don't think it's a great idea to assume that people will invent cookie
records and browser box records and safe browsing records and so on
and so forth.

The schemes that Casey and I proposed used a dedicated new RRTYPE and wildcards
so there's no tree walk, just one deterministic lookup per boundary level.  If
the boundaries for different applications are different, there are flags so each
boundary can say which apps it applies to.

>I rather liked the idea of DMARCbis saying "You need some way to determine
>the Organizational Domain.  One way is with the PSL as described in X, or
>you could do a tree walk as described in Y." 

That seems like a pessimal way to make things interoperate: use one of
an unknown set of algorithms and the other party can't tell which one
you're going to use. If we can't agree that the tree walk is better
than piggybacking on the PSL, I don't see any of the other changes
we're proposing to be worth the effort to republish so we should stop
now and not waste more time.

R's,
John


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

Reply via email to