COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	Dear Mr. Evans:
C00007 ENDMK
C⊗;
Dear Mr. Evans:

	The opportunity to learn what IBM is thinking about with  the
M.I.T.  group  was very valuable to me, and I hope what I have to say
in return will be interesting to you and of use.

	First of all, it is very important to me  that  IBM  make  as
good  computers  as  possible.   The  fact that IBM did not implement
really powerful time-sharing system with the 360  meant  that  no-one
else did either.  Therefore, I will do what I can to help.

	I  am  writing  directly to you rather than to George Raidin,
because I fear that he will find  my  concerns  somewhat  beside  the
point  given the problem he has undertaken to solve.  This is because
I think that the problem he has undertaken to solve, that of creating
a  high  level  machine  language  structure,  is beside the point in
meeting the computer needs of the  next  decade.   In  particular,  I
think   it   is  not  the  main  issue  in  determining  whether  the
applications of computers will grow enough  so  that  the  market  in
dollars will expand rather than contract.

	My  only direct comments on the proposed FS structure are the
following:  Its complication makes  blunders  possible  which  might
force  a  user to get around the intended data and program structures
and build a new structure on top of it that is not really  of  higher
level  but  merely has the slightly different specifications required
for the user needs.  While this is  a  danger,  I  think  it  can  be
avoided by having enough smart people go over the proposed structure.
Incidentally, I think my fellow computer  scientists  will  be  quite
favorably  impressed,  because  the  idea  of  going  to a high level
machine structure is in accordance with public opinion among computer
scientists  in  so  far  as  this  group can be said to have a public
opinion.

	Now for my own opinion as to what the most important issue
is.  This is the creation of a universally usable time-sharing
system.  The problem has the following aspects:

	1. All applications must be programmable in the system with
acceptable efficiency.  This includes both airline reservation systems
and on-line debugging.  Acceptable efficiency does not
require that greater efficiency cannot be obtained by starting from
scratch.  Instead it means that the advantages of the universal system
outweigh the advantages of starting from scratch just as people program
in FORTRAN even though a faster machine language program could
be written.

	2. The programming system and not merely the machine structure
must be long-lived.  In particular, it must be hospitable to new
peripheral equipment including terminals.