Yakov Shafranovich <research(_at_)solidmatrix(_dot_)com> wrote:
Lets step back for a second. We are comparing the SMTP protocol with
other protocols in order to determine which problems the SMTP protocol
has similar to other protocols, and see which solutions applied in other
protocols can help us here.
Exactly. That's the scope of the document I wrote.
The point of contention here seems to be whether SMTP is intended to be
an end-to-end protocol (like a network layer protocol of the OSI model),
or a hop-to-hop protocol (like a link layer protocol).
Almost. We went down a rat-hole of human messaging systems, because
of my contention as to the design intent of SMTP: one to one
communication. If SMTP was intented to be many to many, with high
latencies, it would look like NNTP. If it was intented to be many to
many, with low latencies, it would look like IRC.
That's all I'm going to say on that topic.
So what comes out from this, is that SMTP operates on a hop-by-hop
basis, which in many cases is end-to-end since there is only a single
hop to be done.
Now the question is, what can we, as a RG, learn from this?
The LMAP discussion document (based on Hadmut's comments in RMX),
says that the single hop use of SMTP is a large part of the reason why
spam is so wide-spread. There are 100's of millions of senders, each
sending to only 100's of thousands of recipient MTA's.
How is this behavior helpful to stopping spam?
If that imbalance in the network was addressed, spam would become
significantly more manageable. It wouldn't stop entirely, but having
a blacklist of 100K MTA's is signficantly easier than having
blacklists of millions of IP's.
PHB's suggestion that MTA's have "recommendations" in DNS would
apply here. LMAP would say "yes, example.com meant to originate SMTP
from this IP", and a related protocol would say "example.net thinks
I'm an OK MTA, too." You'd still have to query example.net, so see if
it agrees, but that's fairly simple.
This process builds a web of trust, where MTA's simple don't bother
to listen to spammers.
What comparisons can we draw from this behavior to other protocols?
Many other protocols are only single-hop. Others are nearly always
multi-hop. There are fewer which are a combination of the two, like
SMTP.
If SMTP was one or the other, the spam problem would probably be
easier to solve. It's the combination of factors that makes SMTP
worse than other protocols.
Alan DeKok.
_______________________________________________
Asrg mailing list
Asrg(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/asrg