COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	APPENDIX C - TERMINALS
C00012 ENDMK
C⊗;
APPENDIX C - TERMINALS

	The interactive use of computers requires terminals, and experience
at Stanford, at the office of the Advanced Research Projects Agency and
elsewhere has shown that if terminals are good enough and acessible enough
and connected to a good interactive terminal system, then they will be used
by people whose jobs have not previously been considered to involve computer
use.  For example, the the Artificial Intelligence Laboratory and at ARPA, 
the terminals are used by secretaries and by anyone who needs to write
or to communicate.  In the long run, Stanford will want and will find cost-
effective a terminal system that will give a terminal in his office to almost
every faculty or staff member or graduate student, and undergraduates will
have them in their studies.

	At present, however, neither the terminals nor the computer systems
exist.  Therefore, any plan for terminal use should start out small and
should plan for an eventual very large expansion.  One cannot, however, plan
on using a specific type of terminal during a long expansion period, because
the quality and cost of terminals are still improving rapidly.

	The terminal situation can be summarized as follows:

	1. The cheapest and most widespread terminal is the Model 33 teletype.
It is rugged, reasonably reliable, slow (10 characters per second), and too
noisy to use in an office with other people than the users present.  It also
only allows upper case characters and so is not suitable for reports.  It is
perfectly feasible for use in special terminal rooms for programming.

	2. The most widespread terminal at Stanford is the IBM 2741 based on
the IBM Selectric typewriter.  It is also slow (15 characters per second),
noisy (A computer terminal needs to be quieter than a typewriter, because it
types faster and longer during output than a human typist), and it is dedicated
to half-duplex line-at-a-time operation which is inferior to the more common
full duplex.

	3. There are many teletype compatible terminals that operate at
30 characters per second, are full duplex, quieter than either the
teletype or the 2741, and are cheaper than the 2741 though not as cheap
as the model 33 Teletype.  They also have lower case.  These terminals
dominate the 2741 in every respect except that they are not made by IBM.

	4. Experience has shown that terminals using CRT displays are better
than printing terminals, because they can be much faster, because they are
quiet.  Moreover, some of them can display drawings and pictures, have
arbitrary character sets.  Eventually they will completely displace printing
terminals, although a user should have a printing terminal within walking
distance.  At present,  most display terminals are rather expensive, and
no one terminal is presently available that will displace printing terminals
for all purposes.

This is partly because the original development of display terminals was
for single expensive terminals with elaborate facilities.

	5. At present, the Artificial Intelligence Laboratory operates
a system of sixty display terminals which allows one to be place in each
office.  This system cost about $150,000 plus some internal engineering
that was not accounted for.

	6. The Compter Science Department has decided to acquire a system
of terminals for all of its offices and for a substantial display room for
student use.  It is hoped to have this system installed by Fall 1974.  The
current plan has the following specifications.  There will be 80 keyboard
and display terminals of which 32 can be active simultaneously.  Arbitrary
character sets and pictures can be accomodated since the system will use
a 512 by 512 array of points for each user.  The system will be controlled
by a PDP-11 computer which will be connected to the 360/67 and to the
Artificial Intelligence Laboratory's PDP-10.  The estimated cost of the
system including connection to the computers is $200,000.  It can be expanded
at a cost of $600 per terminal in an office plus $2500 per simultaneously
active terminal.  The major disadvantage of this system is that it requires
coaxial cable connections between the individual terminals and the central
memory.  Reductions that are expected in the cost of memory and computing
should make it possible to put a terminal with these specifications at the
end of a telephone line within three or four years.

	6. At present, an entirely satisfactory display terminal with arbitrary
character set and graphic capability and the ability to be placed at the end
of a telephone line costs about $12,000.  There are a number of such terminals
on the Stanford campus and the proposed interactive system should support them.
Upper case only terminals are in wide use in other places and cost about $3000.
They are basically useful for computer programming and message sending but not
for writing.  In our opinion, they should be supported, but Stanford should not
plan a large system based on them.  There is reason to believe that arbitrary
character set non-graphic terminals will soon be available for less than $1000.
This might be the basis for a system.

	All these facts provide a basis for our recommendation concerning
terminals.  These are:

	1. The campus terminal problem should be considered separately from
the interactive computer problem, because it has a different time scale
for solution and long term decisions cannot yet be made.  Also the demand
for terminals will grow with time and the cost-effectiveness of the
terminals will also grow.

	2. Therefore, the PDP-10 system should be designed to support a
any terminals that have certain minimum capabilities of handling upper case
text.  Graphic terminals require more commitments from the time-sharing
systems and the number of types fully supported may have to be restricted.
A PDP-11 for terminal support is therefore included in the system configuration.