ietf-mxcomp
[Top] [All Lists]

Re: Comments on draft-ietf-marid-core-01 xml use

2004-06-04 11:28:24

Alan DeKok writes:
Has there been a demonstrated need for that extensibility?

MARID seems to be writing a sort of programming language. Mail senders write a program, mail recipients provide an interpreter, run a program provided by the claimed sender, and use the result to decide how to handle the message.

Of course it needs to be extensible. When was there ever a scripting language that didn't need extensibility.

But I question whether that language is a beneficial in the first place. Consider this strawman, which attempts to get rid of the scripting language:

1. A mail recipient looks up _friday._udp.senderdoma.in. (It's Friday today.)

2. If that results in one or more servers, the recipient picks one and sends an UDP packet containing a Friday query to that server. The Friday query contains the MAIL FROM address and SMTP client's IPv4/IPv6 address.

3. The Friday server consider the three facts it has (recipient IP address, sender IP address, claimed sender email address) along with its configuration and other data. It answers "good", "bad" or "uh, dunno".

4. The mail recipient gets a small UDP packet back with the answer.

There. Advantages:

1. No mini-java script language that must be implemented in all mail receivers.
2. The sending domain can implement any policy it wants.
3. No DNS lag for policy changes.
4. No coordination required with different departments when the policy changes. (See earlier thread about email outsourcing, too.)
5. The protocol's much simpler.

Disadvantages:

1. DNS caching doesn't help (much).

Arnt