On Tue, Dec 02, 2003 at 07:41:56PM -0700, Jason R. Mastaler wrote:
> Gerrit Pape <[EMAIL PROTECTED]> writes:
> > Hi, to be able to automatically send a reply to delivery
> > confirmation request messages, the program needs to reliably sort
> > out confirmation request messages from other delivery notifications
> > (bounces).  It also needs to make sure that the confirmation request
> > message is not forged.
> 
> Yup.  There are a few protocols for doing this in the works, and I've
> held off on implementing anything until I see how they flesh out.
> 
> Those I'm aware of are:
> 
> * The Challenge / Response Interworking (CRI) Framework
>   http://www.ietf.org/internet-drafts/draft-irtf-asrg-cri-00.txt
> 
> * The Receptionist protocol
>   http://www.whitelist.com/protocol.html
> 
> One of these might very well become the standard for C/R systems.
> 
> > To make this possible, I suggest that confirmation request messages
> > are created with a special Message-ID.  The Message-ID of the
> > request message is created from the Message-ID of the message that
> > causes the confirmation request, by prepending ``confirm-'', and
> > appending the local host part[0].
> 
> TMDA's confirmation request messages already include the requesting
> Message-ID in an 'In-Reply-To' header, but I suppose this isn't enough
> to distinguish it from other auto-responders, correct?

Yes.  I'm afraid that there may be MTAs out there that add a In-Reply-To
header to bounces.  The difference between confirmation requests and
normal bounces must be definite.

> You'd need the 'confirm-' string in there as well?

Not necessarily, but I think it's a good idea to prepend a keyword here.
The idea of creating a special Message-ID from the original Message-ID
may be used for other types of delivery notifications also, and a
different keyword could be used then.

> > The latest qconfirm version does this, and it would be nice if tmda
> > and ask could do this also.  Presuming that confirmation request

> The problem here is at least twofold as I see it:
> 
> 1) Making this an optional feature wouldn't help, as there isn't any
> incentive for the TMDA user to turn it on.  It would have to be
> enabled by default in order to be useful.
Yes.

> 2) I don't want to implement something that only works between TMDA,
> qconfirm, and ASK.  I'd like a method that is compatible with every
> C/R system out there (see above proposed standards).  So, I'd hate to
> implement this, and then have to revoke it later in favor of something
> else.  And yes, the Message-ID is involved in other proposals, so
> there is the possibility for an incompatibility.

I don't think that there's a possibility for an incompatibility, that's
why I now implemented the Message-ID trick.  The ``Receptionist
protocol'' you mention above uses the Message-ID as identifier, but
doesn't require the Message-ID to be special.  Current standards only
require the Message-ID to be a unique identifier for this particular
mail message, there's no restriction on how it's created.  Creating the
Message-ID for confirmation requests as I suggest doesn't break this, so
it doesn't interfere with the proposed standards.

> What are your thoughts on these issues?  

I think it'll take a long time until one of the proposed standards is
accepted, and since the special usage of the Message-ID does nothing
which could cause trouble for such proposed standards, it can be
deployed right now.

> Do you have plans to push your proposal as something all C/R systems
> should adopt?

I would like to see it adopted, but I'm not the man doing the pushing.
I could write up a technical document on how it works, but my english
skills are not good enough to push it, and convince other people.  It
would consume a lot of my time.

Regards, Gerrit.
-- 
Open projects at http://smarden.org/pape/.
_________________________________________________
tmda-workers mailing list ([EMAIL PROTECTED])
http://tmda.net/lists/listinfo/tmda-workers

Reply via email to