At 03:45 PM 1/14/2008 +0100, Alessandro Vesely wrote:
David MacQuigg wrote:
Starting a new thread, as there seems to be some serious
interest in this topic.
I've created http://www.openspf.org/Community/Glossary and
added a few term's tentative definitions. Please feel
free to amend/expand that at any time.
I'll be happy to contribute to the this page, but first we really need to
resolve some basic differences over the definition of Border. Here is what I
would suggest: http://open-mail.org/Definitions#Border_MTA I am also
considering some changes in my page to better conform with our discussions
here. As you pointed out, the Receiver really should be linked directly to the
Transmitter.
As Frank said, our terminology reflects our "ideology". SPF was originally
designed with the understanding that there was a well-defined border, and
forwarders were only "edge cases". Now there is more emphasis on forwarding.
Crocker seems to have the opposite point-of-view, that a typical mail transfer
involves numerous, unrelated ADMDs.
I think these are just two views of the same world. We can certainly add
"secondary borders" to our diagram, and Crocker can certainly draw a double
line between two of his ADMDs. The difference is really just emphasis. Folks
favoring IP-based authentication methods tend to think of a single border.
Folks favoring signature-based methods tend to think of numerous borders.
I would like to hear more from the other members of this community. Is there a
consensus?
Here is what we have so far for a Basic Mail Handling System.
/
Sender(s) --> Transmitter--> / --> Receiver --> Forwarder(s) --> MDA -->
Recipient
/
Border
This is mostly an Administrative Level description, although
the terms Transmitter, Receiver, and MDA are borrowed from
the Machine Level description. We could use phrases like
Receiving ADMD, or Mail Distribution Service, if the
distinction is important. An MDA might handle such functions
as mail storage, webmail, POP and IMAP access, virus removal,
etc., and it seldom matters whether it is one machine or
many.
A minimum mail system includes at least an MSA (Mail
Submission Agent), a Transmitting MTA, a Receiving MTA, and
an MDA. The MSA and Transmitter could be one machine, and
the Receiver and MDA could be another. There are always at
least two machines and a Border.
If we focus on a single transaction, we can say there are
_exactly_ two (possibly coincident, e.g. if using localhost)
machines and one border between them. On the next hop, the
previous receiver becomes a transmitter (in a generic sense)
and we have a new border that defines its relationship with
the new receiver.
The difficulty stems from the need to draw multi-hops
diagrams that are general enough for the discussion at
hand.
I think adding one forwarder is sufficiently general. If we can handle one, we
can handle many.
Borders are particularly insidious for SPF. "ADMD" has
different meanings when its administrative management
affects DNS, IPs assignment, or SMTP configuration. In
addition, large organizations are partitioned into internal
subunits, which may wish to carry out SPF checking one
against the other. I don't think we can pull out any useful
topology from that.
Can we at least sub-divide the mess into an MON and an MRN,
with a clearly defined Border? I really don't like the idea
that there are Transit ADMDs, floating in cyberspace, not a
part of either MON or MRN. I would call such networks ORNs
(Open Relay Networks), to put them on the same level as MONs
and MRNs. I don't think they belong in any legitimate mail
system.
We can add terms but we don't need to, until they don't enter
the discussion. The Basic Mail Handling System diagram above
may be complicated with the addition of a second "target" MTA
that the forwarder forwards to, e.g.
/
... --> MDA --> Transmitter2 --> / -->Receiver2 --> etc...
/
Border2
I would change "MDA" to "Receiver1", since MDA means Mail Distribution Agent.
The MDA really should be at the end of the line.
Those Transit ADMDs out there are difficult to characterize.
"Open Relay Networks" is a bit crude: An open relay certainly
results in a transit ADMDs, but most of them are not actually
open relays, in the classic sense of the term.
Someone has configured a forwarding mechanism, supplying a
target email address along with any credentials required for
fulfilling that task. "The person concerned", as we may call
the owner of the target mailbox, is not necessarily the same
person as "the perpetrator" who materially carried out the
configuration task. Are such terms needed?
I don't understand these two roles, and how they differ from the role of
Recipient as we have it currently defined. Perhaps an example would help.
I hope there won't be any "Forwarding Mail Transfer Protocol
(FMTP)" as I don't like protocol bloat. However, something
needs to be defined to make forwarding possible in the face
of spam.
Maybe we could define different classes of mail: First Class, Bulk, and
Forwarded. There could be a special blacklist for spammers who try to
masquerade as forwarders. But I digress. Let's finish our discussion of Mail
System Terminology.
-- Dave
-------------------------------------------
Sender Policy Framework: http://www.openspf.org
Archives: http://v2.listbox.com/member/archive/735/=now
RSS Feed: http://v2.listbox.com/member/archive/rss/735/
Modify Your Subscription:
http://v2.listbox.com/member/?member_id=2183229&id_secret=85679465-99265d
Powered by Listbox: http://www.listbox.com