Re: Address to DNS mapping

Hadmut Danisch wrote:
> 
> Gentlemen,
> 
> I'd like to raise another question, which everybody silently
> passed in the drafts (including myself in my earlier drafts).
> 
> Ignore for a moment, what and how we do the lookup. I'd like
> to focus on _where_ we do the lookup.
> 
> When we receive a MAIL FROM: (or HELO) command, it's argument
> is some random text string. Whether we do RMX, DRIP, SPF, all need to
> do a DNS lookup, and we all assumed to just cut the domain part from
> the address and use it as the place where to look at in the DNS.
> 
> I don't think anymore that it is quite that easy:
> 
> - We need to cope with empty sender addresses (coming with
>   error messages),

And (fake) bounce mail.

> - sender addresses without domain

This is only be valid for local (submission) sessions. Many (most)
MTAs currently require a valid, or at least valid looking, FQDN in
the sender address when accepting messages from outside their
administrative realm.

> - empty domain

Isn't this the same as the previous one?

> - malformed addresses (we need a syntax)

If the inputs are malformed, is it worth continuing the SMTP
session?

> - Should we cut off the name part? Why not allow lookup of the
>   full address? And if so, how to avoid ambigousity?

This has some nasty implementation issues associated with it. It
may make a nice extension when all the domain lookup issues are
worked out, though.

> - What about encoding?

Of what.


Raymond S Brand