COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	Dear Steve:
C00010 ENDMK
C⊗;
Dear Steve:

	Here are some reactions to your communications.

	1.  We were  not  expecting to  spend  any of  our  presently
budgeted  money on Foonly.   As you recall money  budgeted for Foonly
in our previous  budget was taken away.   If we  spend such money  on
Foonly, then we  will have to curtail other activities.   Secondly, I
don't  believe  we  have spent  anything  like 1/3  of  our  money on
computing.   Out  of a  total  of $10,000,00  since the  Stanford  AI
Project  started,  I  think  we  have  spent about  $1.6  million  on
computer hardware.  I  will get the  figures on hardware  maintenance
and software maintenance.   Do you know what software  development if
any  goes into this  1/3 figure?   Compared to almost  all your other
clients, I think  we have  been economical on  computing, because  we
never  had an  SDS 940  that had  to be  replaced, we  have only  one
system,  and much of the  hardware is seven years old.   We will come
up with  more accurate  figures. I  think  ARPA should  pay the  full
expenses of whatever plan you  choose. In any case our economical use
of computers that permitted us to get more computing per dollar  than
other projects should not  be rewarded by making us  tax our research
with average costs based on less economical policies.

	Jack Holloway points  out to me that we were planning to have
our present system programming staff  do the conversion to TENEX.   I
guess we can do that without charging it to Foonly.

	2. With  regard to  Stanford's plan to  get a  display system
for  the Computer Science Department,  I am in charge  of it not Vint
although some  change later  may occur.   The fundamental  difference
between Yale type  systems and the ones proposed by  Xerox and ISI is
in the  cost goal.    They are  willing to  have  you pay  $6000  per
terminal  and we  are shooting  for  a system  costing  $200K for  64
terminals  including display management  minicomputer and connections
to the  AI Lab and  360/67.   The marginal  cost per  terminal in  an
office will  be $600.  In  my opinion,   cost is a critical  issue in
the  mass switch from  teletypes to displays and  putting displays in
all offices.   I  don't think  ARPA would have  agreed to  buy us  60
displays at  $6000 each, and we  are now discovering  that some users
need more than  one display at  a time.   Our Datadisc  system of  60
displays cost  you about $150,000.   I  think that the  organizations
that  are pushing expensive  displays without  having experience with
an all display time-sharing system are underestimating the  number of
offices  in  their  places  that should  be  provided  with  displays
especially if office automation continues.

	The  other  differences of  opinion concerning  the following
questions:
	 a. Are 4000 characters required or will the  3200 I now have
on my screen do?

	  b.  Are  super  characters  required or  are  ours  legible
enough?

	 c. Should we standardize a  few fonts or should we go  right
now to arbitrary  character sets?  I  favor the latter and  point out
that   scientific  text  typically  contains   mostly  latin  letters
sprinkled with a random collection of special characters.

	 d.  Is  it important  to  have  raster display,  and  should
displays be compatible  with standard TV (for example,  the TV's used
in vision research)?

	My  own  opinion  is that  ARPA  should let  at  least  a few
flowers bloom  if not  100 and  should  water some  of them.   It  is
imprudent  to set  standards for  display based  time-sharing systems
when only one  is in operation,  namely ours, especially  if you  are
not willing to take advice from that installation.

	3. I  will  have someone  check out  our FTP  with regard  to
sending  files from  us  to a  TENEX  system.   I know  it  works for
transmitting files from TENECes to  us, because I used it to  extract
the ISI display specs  which I haven't had time to read  yet.  I also
know  it works in both  directions between us and  ITS.  The proposed
display system  for the  Stanford Computer Science  Department is  in
the file CSDDIS. PRO[ESS,JMC] in our system.

	I  agree that  there should  be a  common protocol  for using
displays through the net - at least a minimum standard, but I am  not
ready to propose ours,  because we expect to make  major improvements
in connection  with our transition to TENEX.   We are willing to take
part in committee work  aimed at that  end provided the total  number
of committees  concerning all subjects  that we take part  in remains
reasonable.