COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	APPENDIX C
C00009 ENDMK
C⊗;
APPENDIX C

	Because  Stanford  seems to  have  an  extreme  psychological
reliance on  IBM for computing technology, it  seems necessary to try
to document the statement  that IBM is  behind in the development  of
interactive computing.   This might  seem difficult, because  IBM and
its  users have built hundreds  of interactive systems,   and even if
they were all  bad, it would  take ten years  to prove it and  no-one
would read  the proof.   However,  the  main problem arises  from the
fact that  IBM does  not do  interactive computing  within a  general
interactive operating  system.   It has  a few primarily  interactive
systems,   but  it mainly  advocates  doing interactive  computing by
partitioning what is  basically a batch processing  system.  This  is
very inefficient  and leads  to monsters like  the Stone  and Webster
system  described  in  the  attached  article  xeroxed  from the  IBM
Computing Report for fall 1973. Please note the  following aspects of
the Stone and Webster system:


	1.   Some terminals are apparently dedicated  to text editing
and some to other functions.

	2.  The  system is apparently  partitioned into text  editing
and other functions.

	3.   The  text  editing program  is  apparently not  used  to
prepare programs but only reports.

	4.    An additional  370/158  is  to be  procured  to  get 60
terminals of text editing out of  the main 370/165.  If this  370/158
costs the  same as Stanford's  recently procured 158, its  costs come
to xxx per terminal per month.

	5.   The  text editor  is special  rather than a  general IBM
service program.

	6. The subtitle's brag  of "two powerful capabilities  in one
on-line  system" is a  far cry  from Stanford's requirements  and the
capability achieved with  PDP-10 systems.   In either  of the  PDP-10
systems now on  the Stanford campus,   a text editor is  just another
user  program  requiring no  special  privileges.   New  ones  can be
introduced at the  whim of the user,   and the  system will allow  as
many  such capabilities  as are  programmed.   Standard text  editors
exist  that  have all  the  facilities  mentioned in  the  IBM paper.
Moreover,    these text  editors  could  handle  many  more  than  60
terminals  on a  computer  much less  powerful and  expensive  than a
370/158 without having  the time-sharing system  rewritten for  their
benefit.

	The fact that IBM  brags about a system of  which they should
be ashamed  is another indication that they  are retarded in the area
of interactive computing.   However,  with  customers like Stone  and
Webster  who can  afford such  inefficient systems,   IBM  has little
incentive  to do better.   Stanford cannot afford the  cost of such a
system and requires much greater flexibility.

	It could be argued  that the Stanford Center  for Information
Processing  could build a  much better  system than  that out  of IBM
hardware.    Indeed   it  could  and   has,    but   there  are   two
counter-arguments. First, the capabilities of  SCIP in that direction
are calibrated  by WYLBUR.  WYLBUR is a partial  success,  but is far
from giving its users the facilities given on PDP-10's.   Second, any
substantial deviation  from IBM systems  will lose the  advantages of
compatibility  with  IBM software  which  is  built to  operate  in a
basically batch processing  system.   The usual way  out of the  mess
created by the fact that  IBM software is batch oriented is to divide
memory into  partitions  within  which  subsystems  can  be  written.
Unfortunately, this  causes a  rigid division  of the facilities  the
machine into  subsystems that have difficulty communicating with each
other.  It is  also a major decision,   often involving the  purchase
of new  hardware, to make  a new subsystem.   Stanford has had  to go
this  way  on  the 158  in  order to  accomodate  ACME,  the hospital
accounting, and administrative accounting on the same machine.