In order to  give more concrete arguments in favour  of registering local
or  national gateways  in DOMAIN  NAMES, Glenn  Malling and  I have  been
making  a  simple  test  with  the   Novell  list  at  SUVM  (about  1200
recipients). We  have subscribed an account  on a machine which  does not
have  a  local/national gateway  registered  but  instead relies  on  the
INTERBIT service under its BITNET  address, its regular Internet address,
and a %-routed Internet address  which simulates what would have happened
if the  host had  registered a  local gateway.  That is,  we have  made a
comparison of delivery time between:
 
1. BITNET all the way (using DISTRIBUTE).
 
2. Internet all the way (DISTRIBUTE not used - "bandwidth is plentiful").
 
3. BITNET to the local/national gateway (using DISTRIBUTE), then SMTP for
   the final delivery.
 
The test was  made both with a mainframe at  Penn State University, which
has excellent connectivity  and availability, and a  mainframe in Poland,
with  32k  connectivity to  Stockholm  and  64k  to  Wien. Poland  is  an
interesting case because it is by far the best connected eastern country;
that is, if any  eastern country can afford not to  use DISTRIBUTE, it is
Poland.
 
The test was made on a working day (tuesday - monday was a holiday in the
US) and mostly during working hours (the polish test had to extend during
the evening  so that  we could  get figures  we can  compare). It  is not
possible  to easily  calculate the  exact  delivery time  as the  various
clocks on the way are not synchronized  and the error was on the order of
the BITNET  delivery time.  Instead, the  difference between  the arrival
time of the first file in the "triplet" and that of the other 2 files was
calculated, and then an average was calculated. That is, the figures show
the extra delay as compared to the fastest delivery method.
 
+----------+-----------------------+
| Delivery |      Extra delay      |
| method   | Penn State    Poland  |
+----------+------------+----------+
| BITNET   |  Fastest   | Fastest  |
| Mixed    |  42 min 41 | 1 h 45   |
| Internet |   3h 7 min | 3 h 20   |
+----------+------------+----------+
 
BITNET delivery  for is on  the order of  10-15 minutes from  the 'Date:'
stamp  in the  message to  the  arrival time,  about 5  minutes more  for
Poland. Note that in  the case of Poland it was  not possible to reliably
emulate mixed delivery:  the so-called "local" gateway that  was used was
in fact in the US, but the message had the advantage of not being batched
with hundreds of other recipients (many  of them MX'ed) which flooded the
name servers.  Finally, these figures  only apply to  high-volume mailing
lists.  The delays  are partly  due to  the continuous  flow of  messages
(about 100-150 a day).
 
The  notebooks   are  available  if   people  want  to   double-check  my
observations, this  was a 30  minutes programming project. Anyway  I hope
this  will convince  Poland and  Italy to  define a  national gateway  in
BITEARN NODES.
 
  Eric