My previous posting gave a simple example of a node running MAILER
and two other nearby nodes.  That example was too simple to show
that the MAILER PROFILE would need to be different for each MAILER
in much the same way that RSCS route tables are different at each node.
A more realistic example requires 3 nodes running Mailer, each with
some non-Mailer neighbor nodes.
 
              A -- a3 ---- B -- b2 ---- c1 -- c2
            / |            |            |
          a1  a2           b1           C
 
In this example, A, B, and C run MAILER, while a1, b1, etc. do not.
Note that the Mailer nodes need not be on the "main backbone" of
the network -- just fairly near it.  With this topology, A's
MAILER PROFILE would relay mail for any non-local node to MAILER@B
using a BSMTP envelope.  (In this context, "non-local" means any
node except the "A-nodes", a1, a2, and A.)  B's Profile would direct
mail for any A-node to MAILER@A and for all the C-nodes to [log in to unmask]
Finally, C's Profile would relay all non-local mail to [log in to unmask]
 
The main point to note is that to implement a "distribute" function
using Mailers, what needs to be done is to change each Mailer's
profile so that it relays all non-local mail to the Mailers nearest
itself along the appropriate branches of the network.  Thus, Profiles
would no longer be essentially constant throughout the network.
Instead, each Mailer would have a customized Profile very much as
each RSCS or Jnet node has a customized routing file that varies
with one's position within the network.  At the moment, no tool
exists to create such "distribute"-type MAILER PROFILES.
 
--  John Halperin  <[log in to unmask]>
    Stanford Linear Accelerator Center