I am trying to understand the model, which I thought I came around to
realizing was a domain-based signature primarily managed by MTAs. If
I've got that wrong then where did I lose it?
It is indeed a domain-based signature primarily managed by MTAs. It
sounds like the problem is that the rest of us are assuming that it
has to interoperate with the existing e-mail world. That is, if one
of us implements MASS and the other doesn't, and I send you mail or
you send me mail, the result must not be significantly worse than what
happens now. Also, "managed by MTAs" does not mean the same thing as
"managed per-hop."
The majority of existing MTAs DO NOT support MIME in any form. They
pass message bodies through verbatim. MIME is cleverly defined so
that MIME MUAs can interoperate via MTAs that know nothing about MIME
so long as the MTA doesn't mangle the message body too much. That's a
major feature. We cannot assume that MTAs will interpret MIME now,
and we cannot assume that they will interpret MIME in the future. All
we can assume is that they just barely understand RFC 822.
We certainly hope that a lot of MTAs will recognize and interpret
whatever signature goo MASS adds to a message, but we also have to be
sure that the MTAs that don't will still work. That's why MIME
encapsulation, or anything else that requires that the recipient MTA
deal with anything beyond 822 is a non-starter.
For per-hop stuff it's possible to use ESMTP to see what the next hop
handles, but I expect there will be a lot of MASS-aware MTAs behind
dumb gateways, so anything that requires that all of the hops handle
MASS will be at a significant disadvantage compared to something that
will pass through existing dumb relays.
This is also why it would be nice (albeit not crucial) if MUAs could
sign or verify in some circumstances, e.g., a roaming user sending
mail through a non-MASS aware MTA. I don't think it would be a good
idea to demand that signatures survive any possible MTA mangling,
since that leads down the road that MIME took, but we've all seen that
a lot of MTAs pass messages through with little or no change, and a
signature that could survive that would definitely be useful.
Regards,
John Levine, johnl(_at_)taugh(_dot_)com, Taughannock Networks, Trumansburg NY
http://www.taugh.com