On Sun, Jun 14, 2015 at 10:40:15AM -0400, Paul Wouters wrote:
> >As I'm fairly sure I described in detail before, base32 provides the
> >option of reversing the encoding at the server, looking up the local
> >part using whatever fuzzy matching the server wants to use, and
> >sending an appropriate response.
>
> This pretends that the server can read the human mind of the sender.
No, it assumes that the server has access to the same canonicalization
and aliasing data as the SMTP server that processes mail for the
domain. If no such data is available then only exact matches will
return results. With reversible encodings, the server might use
the DNS protocol, but use something fancier than exact match to
locate the right records. Because it is the server for the target
domain, it might the canonical object corresponding to a given
lookup key.
One disadvantage of base32 is that it imposes new limits on the
length of the email address localpart, unless splitting into multiple
labels is introduced to handle longer inputs. Hashing imposes no
such limits.
However, we're then still left with the problem of dealing with
certificates and/or PGP keys that don't match (via the enclosed
user identifiers) the email correspondent that the user asked for.
That requires new code in user-agents to bind additional addresses
to an identity based on authenticated responses that return a key
for Y when you asked for X.
By now many of you have likely noticed that I'm on the fence on
this issue. I have sympathies for both points of view.
--
Viktor.
_______________________________________________
dane mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dane
