Re: the introduction problem, was Email and reputation (was Re: Service outages planned for April 25)

[Phillip Hallam-Baker <[e-mail]>]

On Sat, May 14, 2022 at 1:14 PM John Levine <[e-mail]> wrote:

> It appears that Phillip Hallam-Baker <[e-mail]> said:
> >-=-=-=-=-=-
> >
> >Since y'all are claiming this problem is impossible, I want the glittering
> >prizes if my proposal turns out to work.
>
> I believe that Mesh does what you say, but it's not going to solve the
> introduction
> problem because it's fundamentally not a technical problem.  There is one
> set of
> people who I do want to hear from, a second sent I don't want to hear
> from, and
> a much larger set where I don't know.  People shift among those three
> groups
> unpredictably.
>

My goal is to sufficiently solve the problem so that I spend a negligible
amount of time dealing with unwanted communications.

You keep setting up this binary success/failure.


>People can and will abuse any messaging modality but for the point of view
> >of spam control, I would be MUCH MUCH MUCH MUCH MUCH MUCH MUCH MUCH MUCH
> >MUCH better off is the only folder I was receiving communications from
> >unknown parties was my contact requests folder.
>
> Well, OK, but I can do that with procmail.   Or Boxbe.
>

Not for your telephone calls you can't.

I am not claiming originality. Far from it. I am providing an
infrastructure which allows you to reduce the spam problem to the
introductions problem and then apply any techniques available to the
introductions problem.



> >So for example, I think I would be pretty safe accepting contact requests
> >from:
> >
> >* Anyone who is an Alumni of Southampton, Oxford or MIT
> >* Anyone who has attended an RSA Conference, IETF, OASIS or W3C meeting
> >* Anyone who is an accredited expert witness search agent
> >* Anyone whose validated email address matches one of my SMTP contacts
> >
> >That is going to cover the vast majority of my legitimate contact
> requests.
>
> That's essentially web of trust, give or take the implausibility that
> everyone
> will tag themselves at that level of detail. I get way more expert work
> from
> random lawyers who found me on the web than from search agents.
>

Umm, I am thinking it is more X.509 model but there are some very important
differences.

First off, PGP Web o' Trust was all about validating keys. That is not a
concern in my model. The key binding for @alice is by definition the
callsign registration binding.

The question 'is @alice the person I know as Alice' may or may not be
relevant to me, for purposes of rejecting spam, all I care about is whether
they are an abusive actor.

So no, this is really not Web of Trust even if it has some superficial
resemblances, as does every model of social interactions because that is
what Web o' Trust is an abstraction of.




> >* Anyone with an introduction from someone I have authorized to give
> >introductions
>
> That's exactly web of trust, and we have seen why that doesn't scale,
> because your contacts' preferences aren't yours.  ("Gee, he seemed so
> nice and it would have been rude to refuse.")
>

Not really, not least because Web o' Trust was never much more than an
aspirational goal in the code.

This is a fairly old analysis now, I have moved on, but it explains why my
hybrid model is superior to Web of Trust alone or PKIX alone and quantifies
it:

draft-hallambaker-mesh-trust-09 - Mathematical Mesh 3.0 Part X: The Trust
Mesh (ietf.org)
<https://datatracker.ietf.org/doc/draft-hallambaker-mesh-trust/>


> Just as important, people and entities change.  I buy a widget from
> someone,
> and I give them an address so they can send me a receipt and tracking info.
> Then the week after the widget arrives, they start sending this week's
> specials
> or even worse they share my address with their Treasured Marketing
> Partners.
>

As I said, I am @PHB and that is public. I am not relying on the secrecy of
my address. So they share it with their marketing partners, but those
partners are not in my acceptance list so the most they can do is to send a
contact request.

I get a lot of messages from NewEgg, but I don't read any of them except
the ones with order confirmations. And my goal is that the order
confirmations, shipping notices etc. integrate to my assets management so
those are going to be easy enough to identify and separate.

>
> PS: Madonna will solve her contact problem the same way she does now, by
> paying
> someone to sort through her mail.  For some problems the most effective
> solution
> is to throw money at it.
>

Which is pretty much what I said she would be doing.