Of course, there's already one application-level messaging
protocol that relies extensively on arranged peerings: Usenet.
Usenet doesn't rely on a *full* N-way mesh of arranged peerings,
it relies instead on a core of fairly well interconnected
"backbone" or "core" news sites who've agreed to do feeds with
each other, as well as to feed downstream "leaf nodes" (either
on a for-fee commercial basis, or gratis as part of a regional
consortia or whatever).
To receive traffic or originate traffic, a leaf node doesn't
need to peer with every other news server, it just needs to do
feeds with a couple of upstream core sites to insure that it has
reasonable coverage and redundancy.
Spam isn't much of a problem on Usenet anymore because peers who
tend to have spam issues tend to clean them up or get depeered
or shunned...
There's no reason why one couldn't build a comparable model
for mail, with the SMTP speciality service provider offering
"SMTP transit" to a base of trusted customers. This comparatively
small number of SMTP speciality provider would then maintain
good relations ("peerings") with the comparatively small
number of major ISPs. Oh wait -- there are a variety of folks
who are already specializing in doing that sort of thing --
it's just that most folks don't need to buy that sort of
service (yet).
Regards,
Joe St Sauver (joe@uoregon.edu)
University of Oregon Computing Center
There's no reason why one couldn't build a comparable model for mail, with the SMTP speciality service provider offering "SMTP transit" to a base of trusted customers. This comparatively small number of SMTP speciality provider would then maintain good relations ("peerings") with the comparatively small number of major ISPs. Oh wait -- there are a variety of folks who are already specializing in doing that sort of thing -- it's just that most folks don't need to buy that sort of service (yet).
While this could work, we are mixing a format and content type that is not security sensitive and is used to carry point to multipoint messages (forums?) and media (NNTP), with a format and content type that is highly sensitive, and is generally used to carry point-to-point communications which may contain things like personal or financial information (SMTP).
I am not sure any level of security would make me feel good about passing my emails through a 'peering .. core' of SMTP relays.
However, if we do go in this direction, I plan on firing up my old copies of BinkleyTerm. FIDO and NetMail may be a good place to start 
(Did I just date myself?)
I am not sure any level of security would make me feel good about
passing
my emails through a 'peering .. core' of SMTP relays.
However, if we do go in this direction, I plan on firing up my old
copies
of BinkleyTerm. FIDO and NetMail may be a good place to start
Back in the day, there were users of Fido technology networks
who were concerned with email privacy. They applied a technology
called PGP to secure the contents of their messages.
Interestingly enough, even though the open Internet email
system doesn't normally relay email through other people's
servers, many people still use PGP and its descendants to
secure their email content. Could it be that some people
don't trust routers to deliver their port 25 packets to
the destination without diverting a copy to prying eyes?
Surely people would never configure route maps to do
such a thing?
And let's not even begin to discuss the ways in which
widespread use of crypto technology in email systems could
allow for things like non-repudiable signatures.
And let's not even look at a widespread email alternative
that is in use on the Internet today and which is slowly
gathering features that make it look more and more like
email version two. Imagine what will happen when IM networks
take the same step that email networks took in the early
90's and allow for general interconnection.
--Michael Dillon
Back in the day, there were users of Fido technology networks
who were concerned with email privacy. They applied a technology
called PGP to secure the contents of their messages.
Folks,
We are talking about changing an existing system -- one with an installed base
of perhaps 1B users, but it's fine if we instead use the much lower number of
some "mere" hundreds of thousands of servers.
For such situations, it is considered essential to find a way to obtain
incremental utility, for incremental adoption. It is also considered essential
to minimize the critical dependencies for adoption.
It is considered particularly risky to have one strategic adoption decision
depend upon another, especially when the second has a history of more than 10
years of failing to gain major adoption (or rather, use.)
We could go into a long and painful discussion about the reasons these lessons
are valid, but the reality is that they are not all that difficult to
appreciate, if one looks at the process of obtaining global adoption in a
voluntary environment.
d/