My take-away from the meeting today is this:

- the documents talk about solutions instead of problems.

- perhaps this is because they use terms from a particular design,
  rather than generic terms

- knowing WHO to talk to is a separate problem from HOW to talk to them

  My comments were that the "how" portion should be relatively simple.
I assume a base set of interconnections done via RadSec.  These
connections are provisioned off-line (contracts, legal, phone calls, etc.)

  If A has provisioned a RadSec connection to B, then they share a TLS
session key.  That key can be used to derive additional credentials.

  e.g. A connects to B, and B to C.  If A wants an introduction to C, it
asks B how to find C.  B responds, and gives A and identity based on the
A->B TLS data.  A then connects to C, using the credentials derived from
B.  C can then use it's connection to B in order to validate those
credentials.

  The process can be chained, to allow A to connect to B, then C, then
D, etc.  Each node can decide whether or not it provides an introduction
to another one.  This means that B (above) and be a "registrar", with a
list of members.  It will introduce the members to each other, but not
to anyone else.

  This design is simple, and flat.  It means that each system is nearly
identical.  There are no central registration authorities, except by
prior convention.

  Alan DeKok.
_______________________________________________
abfab mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/abfab

Reply via email to