Re: An e-mail web of trust

Yakov Shafranovich <[email protected]> wrote:
> However, the biggest problem in a web of trust system is scaling

  Only if you have a centralized record of trust.  If you distribute
the system, then the scaling is limited only by local limits on number
of systems trusted.

  e.g. Each domain maintains its own list of trusted systems.

  When a sender wishes to send mail, they establish one or more paths
of trust between sender and recipient.  The recipient can then check
these paths very quickly.

  All of the scaling issues are resolved, and the work of establishing
the path of trust is done by the sender, where it belongs.

  The only problem is that the web of trust now becomes more fragile.
But we already know the net isn't perfect, so I don't see that as a
large problem.

> The real question is would such system be really scalable? I don't know 
> if we can easily answer such question.

  If the system is designed from the start to be a "scale-free" graph,
then the maximum path between participants will be small.  The system
will also be highly distributed, making it more robust.

  e.g. AT&T has business relationships with many other companies.
Those companies have business relationships with others, so the path
between any two tiny businesses may be only 5 elements, and may often
go through AT&T.

  This solution does, however, require the designers of the system to
understand and consciously choose a method of graph partitioning which
has only been known for a few years, and which was coincidentally
discovered by people analyzing the Internet.

  Alan DeKok.