Re: homework, not an experiment, draft-crocker-email-deliveredto

> Following the discussion, I see that the delivery-to is currently used
> (in many cases) to loop detection.
> In such cases it could contain an email address, a token, etc..
> I suspect that at the top of the problems discussed here there is a
> conflict in the header name (deliverd-to:), between the current use
> and the draft.
> I would say that "delivered-to:" is not the best name for the current
> use (loop detection), but... it is there.

And surprisingly difficult to change. Although this type of loop detection is
inherently a local MTA matter, large deployments routinely have lots of
separate MTAs operating in parallel providing a particular service.

Even if you assume that the all the MTAs use the same software - and there
situations where that's not the case - you can't upgrade large numbers of MTAs
simultaneously.

Of course you can argue that this is a minor feature and it won't matter if
it's broken for the duration of an upgrade. But IME the folks whose approval
you need tend not to be very sympathetic to such arguments - understandable
given the costs of failed upgrades and rollbacks.

What this means is that implementations will have to support both
names for a considerable period of time. That's considerably
more expensive than switching the value of a single constant.

				Ned