COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	scienc[w86,jmc]		Letter to Science about the future of computing
C00006 ENDMK
C⊗;
scienc[w86,jmc]		Letter to Science about the future of computing

	The general picture of the future presented in the xxx
issue accords with the views of most university computer scientists,
but it shouldn't be allowed to go without challenge.  Here are some
of its assumptions --- mostly implicit.

	1. While the hardware of computer systems should and will
continue to improve, the operating systems, especially Unix, are
essentially good enough and can be made standard for the forseeable
future.  Unix was originally designed for a PDP-11 computer that 
had 56 kilobytes of memory and is therefore designed to read many
small files in succession, and this makes it slow even when it has
several megabytes.  More seriously, the language of the operating
system is only accidentally a universal programming language.  As a result 
programming it is slow, obscure, error-prone and limited.
While virtual memory allows the programmer to forget
about whether information is in main memory or on disk, he suffers
the consequences in slowness when the operating system reads one
record after another from disk.  Virtual memory has not kept its
promises.
Unix has still other faults, but the main point is that operating
systems are just as much in need of improvement as other aspects
of computing.

	2. The network of networks idea gives a nightmare view of what
might have happened if Theodore N. Vail had not succeeded in
forming A.T.&T.  In fact, he did provide a basis for solving
99 percent of the problem addressed in the article on networks.
Namely, almost all messages and file transfers are small enough
to be economically solved by protocols whereby one computer directly
dials another on the public telephone network.  Even most of the
communication within buildings are most economically handled that
way.  Both the ARPANET and almost all local area networks were
a mistake.