COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	.require "let.pub[let,jmc]" source
C00008 ENDMK
C⊗;
.require "let.pub[let,jmc]" source;
∂LOT Mr. John Leng↓Vice-President
↓Digital Equipment Corporation↓Marlboro, Massachusetts 01752∞
Dear John:

	The KL-10 is now working in our system, and we are very
pleased with it.  It seems to be generally five times faster than
the KA-10 and for some of our applications it seems to be as much
as ten times faster.  I say "seems", because there are now questions
about the accuracy of the previous timings on the KA-10, and we
aren't going to roll back to do them.
We will schedule our dedication ceremony as soon as we can
line up our speakers.

	However, I am mainly writing you about a different matter.
Stanford has decided to create a %2Low Overhead
Time-sharing System%1 (abbreviated LOTS), and I have been appointed
Director; there will also be a manager, and my main interest will
still be the AI Lab.  LOTS has several objectives,
the first of which is to convert use of computing for courses
to a time-sharing mode and to provide new opportunities for undergraduate
and graduate student computing free of charge.
Beyond that, I hope to get students and faculty into the habit
of using the computer to prepare and store term papers, theses and books.

	A DECSystem 2040 with delivery scheduled for November has
been selected, and I think it will work very well.
However, there are two problems with the 2040 whose solution
is required not only for Stanford but also for any large scale use of
the D.E.C. 2040 in educational institutions:

	1. The first is the price of manuals.  We are planning an
extremely low overhead operation; LOTS will have only three employees.
Users will have to teach themselves using the manuals, and students
will buy manuals at the Stanford Bookstore in addition to their
texts.  At the current price of D.E.C. manuals, this is prohibitive.
For example, a student who wants only to program in Fortran should
buy $39.20 worth of manuals in addition to his text, while a student
who wishes to program only in assembly language should buy $79 worth
of manuals.  A student who wishes only to prepare reports using
the editor and print them with RUNOFF need spend a mere $30 on manuals.
The manuals differ in popularity.  The basic %2Getting started with
the DECsystem-20%1 will sell between 500 and 1000 copies per year.  It
is 42 pages long, and its present price is $10.  It should be $1.00 or
at most $2.00 considering its length.  We are talking to Jim Stafford
here about the problem, but I think it requires higher level attention,
because of its importance for widespread student use of the machine.

	2. The second problem is the limitation on number of files and
directories imposed by the operating system.  At present 1600 directories
and 1600 files are allowed, I understand.  I also understand that this
is being increased.  In the AI Lab, we have about 400 user names (some
of which are inactive), perhaps 1000 directories and about 20,000 files.
LOTS will have more users almost from the beginning, perhaps 1000 within
the first year and may grow to several thousand within a few years.  Few
users will have more than one directory, however.  The number of files
may typically be about 10 per user, though some will have a hundred or
more.  Naturally, we will have to buy more disks to accomodate this
demand.
The use of LOTS for preparing and keeping theses, etc. will require
an especially large file system.

	I hope you will be able to help solve these problems, and I
look forward to another mutually satisfactory relationship with D.E.C.
in addition to that involving the AI Lab.

.sgn

cc: James Stafford