COMMENT ⊗   VALID 00004 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	\\M1BASL30\M2BASB30\M3BASI30\M4NGR20
C00015 00003
C00023 00004	for interactive   
C00025 ENDMK
C⊗;
\\M1BASL30;\M2BASB30;\M3BASI30;\M4NGR20;




\F1\CConsiderations Favoring LISP on "Personal" Machines

\JThe computing community is changing. 
Not long ago LISP was a "special purpose list-processing language" used
exclusively by the Artificial Intelligence community, 
requiring very large expensive computing machinery on which to run.
Now LISP is employed to write such varied things as operating systems and 
CAD systems, and is used in several countries to
teach computer science
principles. Furthermore, artificial intelligence people are in high demand
in industry, and  technology is now supplying computing devices that will
support inexpensive high-quality  LISP systems. This note relects on some of the
opportunities that await those who accept the challenge and offer such LISP
systems to the marketplace.

To begin, there  is a substantial
increase in the commercial applications of Artificial Intelligence. Corporations,
such as 
Hewlett-Packard,
Schlumberger, and Texas Instruments, have well-established AI programs.
As a result of this commercialization of AI,
it has become increasingly difficult to locate trained  AI personnel. 
Both the universities and corporations recognize that the success of
industrial AI work
is dependent on an increased pool of people knowledgeable  in AI techniques.
Since the major implementation  language for AI work is LISP, we will see increasing
demand for LISP systems in both the educational and industrial areas.

Two multi-million dollar corporations, 
Symbolics and LMI, have been formed to market versions
of the MIT LISP machine. These computers sell in the $100,000 to $200,000
range and are attractive to those corporations currently involved in AI
research. The long-term market for LISP-based processors is the low-cost
readily accessible product. Such machines will be attractive in two basic
configurations: first, as
"turnkey" applied AI packages were the consumer neither knows nor cares
that LISP is involved; such applications include expert systems, natural 
language interfaces to complex data bases, algebraic manipulation packages
(MACSYMA), and intelligent computer-aided instruction and tutoring.

The second LISP market involves supplying  these low-cost machines 
to those developing  LISP applications.  Besides the flexibility of the
language as a tool for building complex software, professional LISP systems 
involve a "programming environment". This environment includes editors, debuggers,
and language processors, combined into an integrated package that 
the user manipulated through a highly interactive graphics window system.
The primary  examples of this behavior have been developed at Xerox PARC
and MIT; the essential ingredient is a bit-mapped screen that  allows
manipulation of information at "flicker-rate". Examples of systems built in this
style include  Xerox's Smalltalk as well as the LISP systems at MIT.

The Smalltalk system is an elegant example of how one can apply 
graphical techniques to general computation. This system, though lightly
publicized in the last ten years, has recently been released by Xerox to five
manufacturers. The  expected market impact is substantial; from office automation
to education of primary-grade  children, the Smalltalk view of computing
as graphical simulation is powerful and appealing. Their primary medium, like
that of the MIT LISP  machine, is a high resolution (sometimes color) bit-mapped
display.

A simple example of the 
MIT LISP-machine display behavior is presented on the cover of the
March Scientific American. The  program that produced this picture is a small
segment of a LISP  Cube-solving program.  In a more serious vein,
one need only look at  the VLSI magazine, \F3Lambda\F1, to find references
to CAD software written in LISP and utilizing these graphics systems.
In fact, Symbolics and LMI personnel report that a strong element in their
sales involves the attractiveness of this LISP-based VLSI-design software.

Though these specially designed LISP machines are viable now, the
large-scale impact will involve lower-cost  systems implemented on 
general-purpose microprocessor architectures.
Several groups are anticipating the availability of such low cost
"personal"  machines;
the next few paragraphs outline some of these activities.

The Naval Personnel Research and Development Center has indicated substantial
interest in personal LISP machines to support computer-assisted instruction.
They forsee LISP-based machines supporting the training  of personnel throughout
the Navy; they project a need of about 1000 machines per year. 
Other military
and space-related organizations (NASA) are also considering the potential of 
LISP and AI-related  applications.

At MIT Dr. Harold Abelson and Dr. Andrea diSessa  have instituted a program
to revise the undergraduate mathematics and physics curricula using a LISP-based
approach. 
Their project,  co-sponsored
by MIT and ARPA, illustrates two points:
first, a growing educational
market  as a general-purpose  language --LISP 
has long been the major language used in their introductory software course, and
now is to be  used to revitalize the more traditional undergraduate curricula.
Second, ARPA's
sponsorship denotes awareness 
that the growth of AI R&D will require substantial increases in the 
number of AI personnel. The MIT program will do much to improve education and
increase the general awareness of LISP ideas.

Independently,
at Santa Clara University I have instituted a program to introduce 
modern computing ideas into the  undergraduate
 curricula using LISP ideas as the base. This program includes
an interactive programming lab as an integral part of the educational experience.
This pilot study will begin in the spring term and will
be exported to other universities as rapidly as possible; several other universities
have expressed strong interest.
However,
my ultimate goal is to move this integrated program --LISP-based text
and computing laboratory-- into the  high schools to  revitalize their
mathematics and science programs, while giving a more accurate view of modern
computation than that presented in the traditional BASIC model.  

The potential, both in financial and intellectual terms, for an improved
educational offering  is immense. A recent joint NSF and Dept. of Education
task force reported grim predictions for the educational output on the US
technological front; the suggested a Presidential council to guide the
refurbishing of our mathematics and science  programs. Recent issues of Computer
magazine, Communications of the ACM, and BYTE magazine have addressed the same
problems. In  the July 1980 issue of BYTE, Dr. Arthur
Luehrman suggests a billion-dollar equipment market and $50 million annually
in  sales of texts for
computer-related secondary education.

There is increasing awareness that the current uses of computing in education
are inadequate, witness the interest in Smalltalk and LOGO. 
LOGO, as described in Seymour Papert's new book \F2Mindstorms: Computers,
Children, and Powerful Ideas\F1,
has been an exceptional vehicle for teaching
primary-grade students "how to think".  LOGO is Smalltalk's immediate
parent; it was developed at MIT in the late 1960's as a LISP
dialect. LOGO introduced the 
idea of  the "turtle"  and  implemented the rudiments of object-oriented 
programming; both are hallmarks of Smalltalk.
Several
personal computer manufacturers will introduce Smalltalk systems this spring,
and Texas Instruments has  announced LOGO as a product.
Though both of these languages have LISP in their parentage, neither spans
the total range  of problems that LISP addresses. The market for an inexpensive
personal LISP is wide open.

To fill that need I am developing TLC LISP and publishing another LISP book;
this book will address the programming issues that previous offering have
not covered. The combination of quality software and educational materials
will give the developers of complex software the tools they need to apply
LISP effectively.
\.



\CThe Development of TLC-LISP for the Z-8000

\J
To begin, let us outline the background of TLC-LISP.
The LISP Company has planned  this language
to "do for LISP what UCSD did for Pascal";
it is a strong subset of MIT's LISP machine LISP, with a several features added
from other LISP dialects. The documentation gives a complete picture of the
language, but here is a short summary of its features: 
flexible parameter specification,
pioneered by MDL; generalized control structures, including structured iteration,
"escape" mechanisms (CATCH and THROW); full string processing; fixed and 
floating-point arithmetic; 
full property-list  operations; table-driven scanner; generalized
input and output, including "streams"; full LISP macros and "read macros";
comprehensive error control, including type-checking and user-definable error
recovery system; flexible "bibop" allocation; virtualized functions, residing
on mass memory; speedy execution, about 1/3 KA-10 on a 4MHZ Z-80; and a 
large library of LISP functions predefined for the user. This LISP is the only
micro-based implementation that reflects a professional standard; it is quite
popular at MIT, even though its memory size  and interactive facilities are limited.
Our objective is to develop the language into a low-cost professional system.
One segment of this objective involves the dvelopment of a quality 
LISP-based curriculum; the other segment involves the migration of TLC-LISP
from the Z-80 to a larger processor like the Z-8000.

There is an elegant and smooth transition path from the current 
TLC Z-80 LISP to a polished Z-8000 system. 
First, the current TLC-LISP can
be transported to the Z-8000. This  involves, reworking the operating system
interface and smoothing out incompatibilities in the macro facilities
present in the two assemblers. This will result in an immediate, but
modest (64K Byte), LISP on the new machine. A well-defined recoding effort
to replace the Z-80 code with more effective Z-8000 code 
would turn this LISP into an efficient product, suitable for educational
purposes and medium-scale LISP applications. For comparison, such a LISP
would have an effective program size comparable to that of the 7090 --the
AI research vehicle until 1966.
Of course, such a LISP would  have  interactive facilities
 and speed substantially superior to that offered on the 7090.

In parallel with this single-segment effort, a multiple-segment LISP
can be underway. First, we already have operational a multi-bank version
of the Z-80 LISP, offering eight-times the memory capacity of the single
bank LISP. The bank-switching operations are encoded in such a way that
it is a simple task to map them to segment-based operations. This would
give a multi-bank TLC-LISP on the Z-8000 in short order. The recoding
effort of the single-bank version would be directly applicable to this
multi-segment version.

Within the multi-bank version we have included a "window package", similar
to that supporting Xerox's Smalltalk. This window package is currently being
used to develop a display-based programming environment for LISP. Assuming
that the Z-8000 system supports a high-speed video interface, our programming
environment would be transportable to the Z-8000 system.  Finally, we have
been developing a compiler for this LISP; that effort, transported to the Z-8000
would enhance an already attractive personal LISP machine.

Besides the migration of existing code, we should also  enlarge the LISP to
include features that either time or space constraints did 
not allow in the initial Z-80 implementation. In particular, we propose: that
arbitrary-precision arithmetic and a full complement of array facilities
be added to the data structuring facilities; that a "hierarchy system" like
that supporting Smalltalk be added;  that on-line
documentation facilities be added; and finally the continued inclusion in 
TLC-LISP of features in existing LISPs that make sense for the 
next generation LISP. At the implementation level, we also expect to 
make improvements; in particular, more sophisticated garbage collection may
be necessary for the large memories that these new processors support.
One may also improve execution and space performance of LISP code by 
interpreting a more compact representation than the traditional list-structure.

In summary, we offer a elegant, well-defined path from the current
best available micro-processor LISP, to the standard for the next generation
of micro-processors. Given the increased interest in Artificial Intelligence
and therefore in LISP, The LISP Company's strategy of offering the best LISP
in the context of an integrated learning package, will insure a financially and
intellectually fulfilling opportunity to those who wish to participate.
We hope you  will join us.
\.

\←L\→S\←R\-L\/'2;\+L\→L

Yours sincerely,


John R. Allen
The LISP Company
PO Box 487
Redwood Estates, CA
(408) 353-2227
(408) 353-3857

\←S\→L
for interactive   

wics course

santa clara course

navy

68k