ietf-asrg
[Top] [All Lists]

Re: [Asrg] Certs required to send mail

2003-03-25 23:32:38
On Tue, Mar 25, 2003 at 08:23:11PM -0800, Hallam-Baker, Phillip wrote:
Either or questions of this type are usually unhelpful.

Well, I don't want to cause disruption but from experience in other
standards groups with political issues, I now know that it is vitally
important to agree on goals before moving to solutions.

We believe that ad-hoc filtering systems are already negatively
impacting the reliability of email. As does spam itself if left 
unchecked, an email does no good if the reader overlooks it as
spam.

Indeed this is true.  The goal is to get rid of spam, but the goal is
also to seriously minimize mail lost, or so I would put forward.  We should
not create an undependable (or more undependable) mail system in fighting
spam, not if there is a way around that.   We also wish to maintain 
end-to-end designs, protect free speech ... there are many other goals which
we've had various discussions of.

So you filter you loose messages, you don't filter you loose
messages. That is a major problem which can only be addressed
if there is a reliable method of bypassing the filter which in
turn is only acceptable if the method does not create a backdoor
for spam.

We can control what type of messages will be lost, and how many, and in
particular how many will be affected.   For example, if we come up with
a solution where the only mail lost is mailing list mail, and furthermore
mailing lists hosts can take active steps to eliminate that chance, I think
that's a decent outcome, certainly compared to others where it is
unpredictable what mail will be lost.

In fact, predictable loss which can be eliminated may be preferable at
slightly higher loss levels than unpredictable loss, because you can at
least take steps against the former.    With content based filters, for
example, your doctor might be blocked mailing you e-mail about that blue
erection pill -- I'm afraid to name it in an e-mail for reasons you can all
guess :-) -- because spammers often mail about it, but there's no way to
predict.

This is only one of the goals to be balanced however.   Stamps do meet
this goal, in that you can assure non-blockage by adding stamps, and so
do certificates, but of course they have problems meeting other goals.


As for the 'end to end' principle, on what basis is it asserted
that there are only two ends in this transaction? I see at least
four endpoints, the user relationship is quite separate from the
relationship between the enterprises that provide connectivity
which in turn is separate from the relationship between their
employers if it is an enterprise to enterprise transaction.

In a true end to end system, there would be two endpoints.  However other
realities, primarily intermittently connected endpoints, NATs and underpowered
machines have led most people to do E-mail via a 4-point solution.  However,
that's not true for me.  My mailer, mutt, is running on the very same machine
with my SMTP server.  For me to and from other people like me, E-mail is
fully end to end, as it was designed.

While we want the ability to use proxies, mailbox servers and relays to
get around the limitations above, do we wish to mandate it?  That's another
question.

I am not speaking to anybody's specific views here on end to end designs.
And strictly speaking, they are not even a goal, they are a means to some
other goals (fostering innovation, flexibility of change, and yes, protection
from outside control.)  However, the design principle sits as a proxy (ironic,
that) for the collection of goals.

Nonetheless I am concerned with those who trumpeted behind us when we
struck down laws like the CDA and fought other laws which would blame the ISP
for the packets sent by their users.   (Now if ISPs proxy for their users
that can be another story, of course.)   Today an Hoster in the USA shut down a
Canadian web site which hosted pictures of captured U.S. prisoners because
their ISP (Level3) put pressure on them to do so.

We seem to want to apply pressure to blame the ISP when we don't like the
packets, but want to not have the ISP blamed in other circumstances.

Fortunately, spam should not force us into this dilemma.  Spam violates the
two rules midpoints are allowed to care about in the end to end principle,
high unrequested traffic volume.   We can say, "Stop that user he's sending too
much unwanted traffic" to a midpoint, which is all we need to deal with spam.
We should not be able to ask the midpoint to punish a user because of what's
in the packets, because that way lies madness.
_______________________________________________
Asrg mailing list
Asrg(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/asrg