Jim,
What you observed is mail from a list that whose owner has set
Mail-via= Distribute. This means that mail sent to server A will
result in A sending a DISTRIBUTE job being sent to servers B, C, and
D, and single-address mail to those users closest to it. A partitions
the readership into sets which are best reached via B, C, and D, and
then sends just those readers to those servers. When B receives its
share, it repeats the process, possibly forwarding some readers to E
and F (A understands that E and F are further from it than B, and B
had some readers anyway, so A let B handle the re-partitioning). This
is controlled by the PEERS NAMES and LINKSWT FILE files on LISTSERV,
which Eric maintains. The intent is to limit the number of files
transiting a specific link to as few as possible. I agree with you
that, if A has to send to 3 users at X (which is nearest to A), it
would make sense to combine those into one file, but it doesn't, and I
don't feel a need to bash Eric about it, since DISTRIBUTE does a nice
job of handling more distant nodes.
On the "Who else is interested?" point, you never really had a
complete list anyway. LISTSERV would send individual mail to users on
the local system, and gang-of-five mail to remote systems (1 to 5
users per mail file, to reduce header length (thus increasing
readability)). So there might be 9 users at YALEVM reading the list,
but you'll only know about 5 or 4, depending on which group you fall
into.
Ross Patterson
Rutgers University
|