On Tue, 21 Apr 1992 07:49:53, Kathryn Wright <LIBKAT@INDST> wrote:
>Remember when I asked about Internetted mail from INDYCMS being routed to
>UICVM for distribution to INDST?  Well, now it's not UICVM in its Internet
>capacity - it's Princeton.  What's going on?  This looks ridiculous.
>
I sent a 'REVIEW LIBRARY' command to [log in to unmask]
Your subscription is listed as:
 
       [log in to unmask]         Kathryn S Wright
 
A few other subscribers are:
 
          [log in to unmask]        Elizabeth Chisato Uyeki
       [log in to unmask]        Linda L Messersmith
        [log in to unmask]  David Lewis
         [log in to unmask]      C M Kelley
 
but they are not relevant to this discussion, except to say that explicit
(and, IMHO, unnecessary) routings like these impair the efficiency of
LISTSERV (and the network-in-general) in delivering the mail.
 
Then, I sent a 'SHOW ALIAS INDST' command to 'LISTSERV@INDYCMS',
and got the response:
 
 Node INDST has no known alias.
 
So, I sent a 'SHOW NODE INDST' command to LISTSERV, and got:
 
 Information from the VERS9205 BITEARN NODES file:
   :lastup.REP 19910628 someone@INDST
   :machine.IBM 4361
   :nodedesc.Indiana State University Computer Center
   :p_cedwards.Champe Edwards;CCHAMPE@INDST;+1 812 237.2961
   :p_fmcwilli.Floyd McWilliams;CCFLOYD@INDST;+1 812 237.3370
   :p_tann.Toulson Rose Ann;CCCRAT@INDST;+1 812 237.2928
   :routtab.RSCS (NETSERV,CCHAMPE@INDST)
   :servers1.MAILER@INDST(MAIL,PU,M,BSMTP,p_tann)
   :system.VM/SP
   :techinfo.p_cedwards
   :useradm.p_cedwards
   :userinfo.p_fmcwilli
 
Notice that there is no ':internet.INDST.INDSTATE.EDU' tag defined.
Therefore, LISTSERV has no way of determining that the nodeINDST.INDSTATE.EDU
is (directly) accessible by the BITNET link from INDYCMS to INDST.
 
When LISTSERV@INDYCMS uses the mail-distribution technique which is defined
for this specific list, i.e., 'Mail-via= Dist', it takes one copy of the
posting, merges it with a partial list of subscribers, i.e., all the
"up-stream" (closer to UICVM) IDs, and sends one "job" to LISTSERV@UICVM
and asks it to distribute the posting.
 
LISTSERV@UICVM distributes the posting to each of the nearby BITNET nodes.
Then, since LISTSERV@UICVM also doesn't know the equivalence between the BITNET
and the Internet names for INDST, it ships a copy of the posting, and a
smaller subset of the subcriber-IDs, to LISTSERV@PUCC for further processing.
It's obvious that PUCC runs one of the BITNET-to-Internet gateways,
so it sends the posting (over Internet) to INDST.
 
Why did LISTSERV@INDYCMS choose to send a "job" to LISTSERV@UICVM,
and why did LISTSERV@UICVM choose to send a "job" to LISTSERV@PUCC?
The answer is that LISTSERV examines the nodes of all the IDs in the sublist,
and then selects one of the "LISTSERV backbone peers" as being the "most-
efficient" for this specific set of subscribers -- even if some of the
routings are non-optimal, the overall routing **is** optimal.
 
So, what can you do to help?  The most important thing is to arm-twist
the gurus at INDST (names are shown above), and convince them to add that
':internet.VALUE' tag to the BITNET node-definition for INDST.  Then, after
LISTSERV@INDYCMS receives the updated node-entry (at the beginning of each
month), it will recognize that the most-efficient route from INDYCMS to
INDST.INDSTATE.EDU is over the BITNET link to INDST.  Until then, remove
your Internet ID from the list, and add your BITNET ID.