COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.device xgp
C00015 ENDMK
C⊗;
.device xgp
.turn on "%"
.font 1 "ngr25"
.font 2 "ngr30"
.indent 6
.tabbreak
.page frame 59 high 81 wide
.title area footing line 59
.every footing(,%1{page}%*,)
.title area mike line 1
.next page
.place mike
.begin center; select 2
A DISPLAY TERMINAL SYSTEM FOR THE COMPUTER SCIENCE DEPARTMENT
.end
.area henry lines 4 to 57 in 2 columns 38 wide
.place henry
	This  is a plan to equip the Computer Science Department with
a system of keyboard and display terminals for use  in  teaching  and
research.   The  terminal  system  will make accessible to department
faculty and graduate students from their offices the  IBM  360/67  in
the  Computation  Center,  the  PDP-10 in the Artificial Intelligence
Laboratory, and through the PDP-10 any computers on the ARPA  network
that  they  may  have  made  arrangements to use.  This includes, for
example, the ILLIAC 4 at Ames Research Center.  Connections to  other
campus computers can also be arranged.

	In our opinion, such a terminal system together with suitable
hard copy output  in  the  most  important  laboratory  and  teaching
facility  that a computer science department can have.  The interests
of the faculty and students are too varied to be satisfied by any one
computer  since  no  one  computer  will  have  all  the hardware and
software facilities that are of research or teaching interest.

	The plan we are now proposing is based in  part  on  a  study
made  by John McCarthy at M.I.T. in fall 1972 in support of a plan to
get a terminal system for Project MAC and the Artificial Intelligence
Laboratory.  In  this  study,  requests  for  proposals  were sent to
leading firms in the display field and the  replies  were  evaluated.
The  best  plan,  however,  was found to be scheme being developed by
Peter Wiener of Yale University, and our plan is based  on  the  Yale
scheme  which  is  also  being  followed  by  the  M.I.T.  Artificial
Intelligence  Laboratory.   The  Stanford   Artificial   Intelligence
Laboratory  also hopes to upgrade its Data Disk display system to the
new standard.

	The goals that we want to achieve are the following:

	1. A terminal in each office used  by  faculty  and  graduate
students. This requires very low cost terminals.

	2.  A  large  terminal  facility  for  the use of students in
courses. About  half  the  terminals  would  be  available  for  this
purpose.

	3.  Each  terminal  should  be capable of displaying at least
half a page of arbitrary  character  sets  and  also  graphics.   The
character   sets   used   in  programming  and  in  mathematics  have
proliferated to the point where allowing each user to  have  whatever
characters  he wants is the only solution compatible with the goal of
using any computers that are available to our faculty and students.

	4. Printing facilities compatible with the above should  also
be  available but not necessarily with each terminal.  In fact, there
is no present  way  of  making  such  facilities  available  to  each
terminal  but  suitable  public  printers can be had, for example the
Xerox XGP. The cost of  adding  such  a  printer  has  not  yet  been
determined, but there is a strong demand for it, and its cost will be
included in a future version of the plan.

	The scheme proposed has the following characteristics:

	1. The display image for each active  user  is  stored  as  a
512x512 raster in an integrated circuit read-write memory.

	2.  There are 64 terminals, but only 32 memories so that only
half of the terminals can be active simultaneously.  However,  a  low
duty  cycle is to be expected if the terminals are in offices so this
is ok.

	3. The cost to add a terminal to the system  is  about  $500,
and  the  cost  to  add  a  memory  is  about  $2200 now, but a large
reduction is expected in less than two years as 4096 bit memory chips
come  into  production  replacing  the  1024  bit  chips on which the
present prices are based.

	4. The system is controlled by a mini-computer, say a PDP-11,
which  communicates with the host computers (initially the IBM 360/67
and the PDP-10), and  writes  characters  and  pictures  by  directly
addressing  the  display  memories.  The displays are maintained by a
memory port that reads bits from the memories  into  shift  registers
and  transmits  them  together with synchronization signals through a
video switch to the displays over coaxial cable.

	5. Each display is connected to the central  unit  containing
the  memories by an individual cable, and the keyboards are connected
to the keyboard multiplexer by twisted pair unless it turns out to be
feasible  to  transmit  the  keyboard signal back on the same coaxial
cable as is used for the video.

	6. The video switch is a 32x64 electronic crossbar.  It  will
be  a  copy  of  a  unit  now  in  use in the Artificial Intelligence
Laboratory's display system.

	7. In the initial version of the system  all  characters  and
vectors  are  produced by software in the mini-computer.  A PDP-11/45
will take 150 microseconds in the worst case to  write  a  character.
If  the host computers can support a higher rate of character writing
then a special character writer can be added.

	Here are the components of  the  proposed  system  and  their
estimated costs:

.begin
.nofill
.turn on "→"
.group skip 1
1. Integrated circuit memory→$101.4K
(with power supplies and cabinet)
(based on a verbal quote from
Intel received Nov. 1, 1973)

2. Memory interface cards→$16K
control electronics

3. One time engineering cost→5K
to Intel.

4. Keyboards→$14K
(70 keyboards at $200,
based on quote by Microswitch)

5. Monitors→$14K
(70 monitors at $200,
based on quote by Ball Bros.)

6. PDP-11/40 with 16K words→$20K
(D.E.C. published prices)

7. Video switch→$9K
(cost of video switch built by AI Lab)

8. Keyboard multiplexer→$8K

9. Connection to PDP-10→$15K

10. Connection to the IBM 360/67→$10K
(in addition $300 per month rental
for IBM hardware is required)

11. Contingencies→$10K

                        TOTAL→$222.4K

.end

	Not included in these costs are engineering which we hope  to
do  with volunteer labor and wiring costs.  The necessary programming
will  also  be  done  within   the   Computer   Science   Department.
Maintenance  of  the system should cost about $5000 per year counting
student labor and parts costs.

	Perhaps $50K of the costs can be obtained from projects.  The
AI  Laboratory  should  pay  for  its  connection and the cost of the
terminals used by project members on campus.   With  NSF  not  giving
facilities  grants  any  more,  the  outlook  from that source is not
bright.  We believe that the University should pay the main  cost  as
the  major  laboratory  instructional  and  research  facility of the
Department.

	Regrettably, there are a few costs not yet listed above.  The
major one is stringing the cables, and this has not yet been explored.
The cost of the cable itself is not large - 4 cents per foot, but the
cost of stringing it may be significant.  If we figure 500 foot
average cable length, we get $20 per terminal for the cable or $1280
altogether for the bare cost of the coax.  The cost of the keyboard
cable will be less.

	The engineering can be done by Jeff Rubin of the Artificial
Intelligence Laboratory who built a Yale type system at  M.I.T.  

	A recent survey of offices in the Computer Science Department
turned up 57 locations where terminals would be wanted.  I am not
sure whether secretaries were included as they should be according
to AI Lab experience.  This suggests that a somewhat larger system
would be wanted if we are to meet all CSD demand and provide a
substantial terminal room (at least 16 terminals) for students.
Therefore, we recommend $200K as the total cost of the project not
counting any overheads.