Re: users with vanity domains

Hadmut Danisch wrote:
> 
> On Tue, Oct 14, 2003 at 01:53:37PM -0400, Raymond S Brand wrote:
> >
> > I just checked the 03 draft and didn't see how RMX solves the problem. What
> > section is the solution in. I did find discussion of the problem and the
> > suggestion the the POP server could be used for outbound mail but that isn't
> > part of the RMX proposal.
> 
> See section 5.4.1 on page 13.
> 
> RMX has an entry type for domain names, pointing to a host.
> It is up to the host to keep the DNS name up to date, e.g. using
> DynDNS. E.g. if the RMX entry for my domain contains an entry
> host:notebook.danisch.de  and my notebook performs a DynDNS update
> after dialing in, the notebook is permitted to deliver.

Thanks, I was searching for mobile ...

OK, this is a variation of the ``mobile users are required do a DNS update
operation prior to sending mail'' that all of the proposals support in one
fashion or another.

> > > But this is one of the main reasons why I do believe that we need
> > > a client library with some intelligence and cannot use just primitive
> > > DNS querying through rewriting rules.
> >
> > Any standard that is so complicated that we need to supply a client
> > library so that sites will implement it, is not likely to be implemented
> > at enough sites to be useful.
> 
> Dou you believe you're able to give a description of a ruleset which
> can be processed by every MTA? If we leave it up to every MTA
> implementor to interprete the draft in his personal flavour, we won't
> have the non-ambiguity we need.

Then the draft _IS_ ambiguous and needs to be fixed.

> Most MTAs will need a library
> implementation anyway, e.g. all those Firewall MTAs. And as far as
> I know an RFC also requires a reference implementation. If we have to
> provide a reference implementation, we can make this available for
> free.

RFCs do not require reference implementations; STDs are STRONGLY encouraged
to have two independent, interoperable implementations.

> In other words: Describe how you would like to do it without a
> library, and I'll tell you which MTA will not be able to do it.

Produce a document that describes the process in sufficient detail that
anyone can implement it correctly.


Raymond S Brand