Here is a table showing new nodes minus drop outs. Note that CREN charges
are due in July, so naturally this  is when most people drop out. January
(really Christmas) is also a high drop-out month.
 
8701 61       Back when BITNET was doing well
(...)
9107 -8       First negative ever
9108 -20
(...)
9207 2        About same level as 91
9208 -18
(...)
9301 -52      High score
(...)
9307 -38
9308 -26
9309 -45
9310 -7
9311 -45
9312 -19
9401 -88      New high score
9402 -30
9403 -43
9404 -84
9405 -29
9406 -10
9407 -229     New high score
 
Projections based on 9307-9309 behaviour:
9408 -157
9409 -271
 
My personal projections:
9408 -130
9409 -175
Less than 2000 nodes by 9412
 
While it is fashionable to claim that  the problem is due to the death of
proprietary protocols,  the real reason for  these drop outs is  the high
cost of BITNET membership in the US, as evidenced by the fact that people
leave mostly  when the bills  are due. Some  universities pay as  much as
$8,000.00/year for NJE  connectivity. The actual cost is on  the order of
$50-70/node/year,  based on  an  estimate made  by  EARN, which  actually
provides the  BITEARN NODES  generation and coordination  (other networks
collect the  data from  their members,  and ship it  to EARN  for central
processing).  There  has been  an  attempt  at starting  an  independent,
neutral structure to sell just NJE  table entries for that kind of prices
(presumably with a minimum cost of $300/year to cover the increased costs
that would result from dealing  with 1000 individual organizations rather
than  4-5  large ones).  Unfortunately,  CREN  was  able to  enforce  its
monopoly. Due  to downsizing, people are  no longer willing to  pay $4-8k
yearly  for a  service  that should  cost  a few  hundred,  and they  are
dropping  out en  masse.  When a  regional  hub drops  out,  many of  the
downstream nodes  follow. Once BITNET  goes below critical mass,  it will
collapse as there will be no point in being connected any longer.
 
  Eric