On Fri, Sep 27, 2019 at 5:02 PM Valdis Klētnieks
<valdis(_dot_)kletnieks(_at_)vt(_dot_)edu>
wrote:
On Fri, 27 Sep 2019 15:37:19 -0700, David MacQuigg said:
I view the Mailstore as part of the MDA, whether it is actually disks
inside the MDA, or somewhere out in the cloud. Maybe leaving out the
label
"Mailstore" will make the diagram more clear. I don't see any scenario
in
which we need to treat them as separate Agents. The essential
simplification is that the Recipient can view them as one Mailstore and
Delivery Agent.
That simplification is OK if you want the reader to understand the view as
seen
from a Recipient.
The User's view (Author or Recipient) is a good starting point for a
simple, informative model. I've used this model in lectures to students in
a Computer Networks class at U of A, and as a basis for building my own
experimental email system, box67.com. In class, we never get as far as RFC
5598, but I firmly believe that students are better prepared to learn these
details, if they ever need to later in their careers.
I am not worried about a few necessary departures from the terminology and
models in the RFC. CS students are used to a boatload of new terms with
each chapter. What's important at this level is focus on fundamentals and
building a solid foundation of understanding. I don't let students think
that my presentation is the whole story, especially with complex systems
like email. I show them Fig. 5 in RFC 5598 to let them know there is lot
more to learn if they ever get serious about email systems.
However, if you're actually designing a Mailstore or Delivery Agent
that has to scale to large sizes, you're *really* going to want to have
the distinction
clear in your mind.
I've never designed an MDA or a Mailstore, but I believe you. Your example
below is excellent.
True story: My recent employer had a need to set up ITAR-compliant email
for
researchers - and one of the requirements is that the data has to reside
in servers
on US soil. This was a problem for GMail, because we couldn't get Google
to guarantee
that if one of our researchers was in Europe and Google's location service
learned about
it, that it would *not* pre-stage the data in the GMail mailbox to
European servers for
faster access if/when the researcher checked their mail.
So in that case, the delivery agent that drops the *first* copy into the
mailstore is *very*
different from the mailstore itself which is doing replication and moving
from continent to
continent.
We ended up standing up an Exchange server and some NAS storage in a
locked room
with more access controls than the rest of our server room.
I'm curious if this setup conformed to Fig. 5 in RFC 5598.
MDA --> Mailstore --> Recipient
I suspect this is a case where the Mailstore would be "off to the side" in
a Relay-level diagram, and connected via a secure, high-bandwidth line
crossing the international border.
MDA --> Recipient
//
Mailstore
If you slavishly follow the RFC, then the Mailstore has to replicate all
the functions of the Message Delivery Agent, authenticating Recipients,
running POP, IMAP, and HTTP, etc., and will likely be a lot less secure.
_______________________________________________
ietf-smtp mailing list
ietf-smtp(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf-smtp