Skip to main content

Full text of "Interactive Computing newsletters"

See other formats

j fr f i 

The following articles are reprinted solely as items of interest for the independent evaluation by members of The Association of Time-Sharing Users and The 
Association of Small Computer Users. The opinions, statements of fact, and conclusions expressed herein are not those of either Association. 

Time-Sharing ‘Messiah’ 
Victorious at Last 

By Miles Benson 
Special to CW 

Time-sharing is one of those 
things like pregnancy — you ei¬ 
ther have it or you don’t. 

What I mean is, when it comes 
to the use of time-sharing facili¬ 
ties for software development in 
a company, the company either 
does it or it doesn’t. Rarely does 
a company allow programmers 
or projects the option of choosing 
between time-sharing and batch 

Clickety-Clack and Western 
Railroad is one of those nontime¬ 
sharing shops. It is an unwritten 
company policy that “thou shalt 
hold no CRT images before thee, 
and thy testings shall only spew 
forth from a batch printer.” As if, 
parentheticajly, someone had 
said, “If God intended for people 
to talk to computers, he would 
have equipped them with mo¬ 

And Slim Griffen knew that. 
When Slim hired in at Clickety- 
Clack five years ago and was 
given a tour of the computer cen¬ 
ter, his only comment had been, 
“But where is the terminal 
room?” When his manager-escort 
had responded “The what?”, 
Slim’s worst fears had been real¬ 
ized. A computing shop without 
time-sharing, Slim felt, was like a 
Cadillac in a gas crisis — it was 
elegant and impressive, but not 

of much value. 

You see, Slim’s computing ca¬ 
reer had been nurtured at P.Q. 
Aresti U., a favorite haunt of in¬ 
tellectual folk and a veritable 
bastion of belief in time-sharing. 
Basketweaving majors used ter¬ 
minals at Aresti. Janitorial serv¬ 
ices used a computerized man¬ 
power allocation system. And 
even the football coach played 
simulated games via a Decwriter 
in the locker room. 

With that background, you may 
wonder why Slim ever accepted 
the offer from Clickety-Clack. 
Well, there are many perversities 
in human nature. Perhaps the 
worst of them is a missionary at¬ 
titude. Slim really believed God 
was on his side and that time¬ 
sharing would eventually be¬ 
come a way of life on his new 
job. He even saw himself as the 
Messiah on the White Steed (his 
training in comparative religion 
was not up to the standards of his 
computer science, as you can 
see) who would ride in and slay 
the dragon of batch printers. 

It was also true that he deeply 
wanted to live in the Monica 
Beach area along the Northern 
California coast, and Clickety- 
Clack was really the only com¬ 
puting game in town. But that 
practical explanation lacked ro¬ 
mance, and Slim tried not to 
think of it. 

The Messiah biz at Clickety-Clack 
was frustrating, Slim found. No, not 
just frustrating — it was impossible. 
The power structure surrounding 
Slim hadn’t the slightest interest in 
Slim’s terminal-in-every-pot plot. 
And several years went by in which 
his White Steed stood around and 
swished flies with its tail, ate hay 
and did all the other things White 
Steeds are wont to do. But it never 
came near the batch printers. 

But Slim was not a quitter. He still 
knew he was right. And if logic 
would not move the establishment, 
then perhaps being devious would. 

Now it chanced that Clickety- 
Clack was involved in a peculiar 
business at the same moments in 
time in which Slim’s devious plot 
was conceived. Although most of 
the software developed at Clickety- 
Clack was for internal consumption 
only (“sticking to our own tracks” 
was the mildly humorous in-plant 
expression), a decision had been 
made to develop some “end- 
product” software. 

In the course of sticking to its own 
tracks, Clickety-Clack had built a 
fairly nifty train scheduling algo¬ 
rithm. When some of the other rail¬ 
roads began making inquiries about 
the system built on that algorithm 
(it was called Train Automated 
Routing System, or Tars), Clickety- 
Clack computing management 
quickly realized it had a potential 

money-maker on its hands. Its re¬ 
sponse to the other railroads was a 
brief description of the system and 
a promise that a proposal to define 
and implement a generalized Tars 
system for multirailroad use would 
be forthcoming. 

And Slim was assigned to the writ¬ 
ing of that proposal. The plot thick¬ 

A ‘Natural’ 

Slim was a natural to work on the 
Tars proposal He had designed 
most of the system and imple¬ 
mented a fair chunk. And what he 
hadn’t built himself, he had learned 
inside-out. If anyone knew the 
Clickety-Clack-specific parts of 
Tars, the parts that needed redoing, 
it was Slim. 

Well, to get the story moving again, 
Slim dashed off a technically credi¬ 
ble, well-worded, tactically sound 
proposal in record time. It was 
when an audit committee was as¬ 
signed to go over Slim’s work before 
the proposal went out that the thick¬ 
ening plot really almost clotted. 

I was assigned to that audit com¬ 
mittee — along with two Clickety- 
Clack computing managers. I was 

asked to look at the technical de¬ 
tails. The managers, naturally, 
looked at the Big Picture. 

Slim had done a superb job. The 
managers were pleased with the 
Big Picture. The technical details 
were impeccable, And it was only 
on my second reading that I came to 
realize what Slim had done. 

Buried deep in the proposal, in a 
section labeled “Implementation 
Methodology,” was a pass-over- 
quickly sentence that said, “Devel¬ 
opment of the generalized Tars sys¬ 

tem shall be done in a structured 
manner, via a modular design and 
utilizing time-sharing techniques.” 

On first reading, I had zipped past 
it as Proposalized Motherhood. If I 
had quit then, no one else would 
have ever noticed it. Managers sel¬ 
dom read sections which begin with 
the word “implementation” or end 
with “methodology.” That level of 
detail they generally find boring. 
And, of course, Slim knew that. 

Have you got the picture here? 
Slim had slipped into a formal 
Clickety-Clack document a promise 
to use time-sharing. If it went un¬ 
challenged and a contract resulted 
from the proposal, Clickety-Clack’s 
unwritten code had to change. Slim, 
mounted once again on his White 
Steed, had found a back-door ap¬ 
proach to waging his fervent reli¬ 
gious war. 

And I was standing in his path. It 
was my responsibility, as part of the 
audit team, to report my findings. 

But — and here is where the plot 
really crystallizes — there was a 
loophole. Whereas the managers 
reported their findings to manage¬ 
ment, I needed only report my find¬ 
ings to the technologist — Slim. 

I’d always been somewhat sympa¬ 
thetic to Slim’s battle. I’m not sold 
on time-sharing as a way of life, 
mind you, bull’d like to see it tried 
in order to have more data for a bet¬ 
ter decision., 

You see, I love the idea that 
management’s disinterest in the de¬ 
tails of computing technology can 
be used in potent ways against it. I 
think I slid Slim’s devious plot 
through for that reason more than 
any other. 

Reprinted by permission from Computerworld, October 9, 1978, ©Copyright 1978, CW Communications, Inc. 

Members: I find it hard to agree with this author’s conclusions, but I think we should be aware of his arguments. HS 

Blows Against 

the Timesharing Empire-_ _ BY JOHN WALKER 

The development of microcomputers 
has, recapitulated the development of 
large computers of the Fifties and the 
minicomputers of the Sixties. The first 
systems had tiny amounts of memory; 
were operated at the bit level with 
switches and lights; and were built and 
used mostly by people with intimate 
knowledge of hardware. As the price of 
memory declined and an infrastructure 
of software was built, systems advanced 
to the point where they could be used 
by software-oriented people, program¬ 
ming in either BASIC or assembly lan¬ 
guages. The past year has seen the intro¬ 
duction of complete application pack¬ 
ages for microcomputers and turnkey 
microcomputer systems. 

As the declining price of semicon¬ 
ductors continues its asymptotic skid 
towards zero, the products of ingenuity 
from peripheral designers seem never to 
stop. In a couple of years one may be 
able to assemble a system comparable to 
contemporary medium-scale mainframes 
for a price not much more than today’s 
personal computers. Wonderful,but what 
do we do with it? Obviously, because 
we’re repeating history so faithfully, we 
build a timesharing system. This is a 
pretty safe prediction (it being a lot eas¬ 
ier to predict the present than the fu¬ 
ture) because at least three microcom¬ 
puter manufacturers are offering time¬ 
sharing systems already, and undoubt¬ 
edly many more will be thundering to 
the finish line. This is a terrible mistake. 

To understand why timesharing and 
microcomputers should not be mixed, 
we must look at the history of time¬ 
sharing and the reasons behind its de¬ 
velopment and current wide acceptance 
in the computer community. From al¬ 
most inception, computers, it was real¬ 
ized, could be used most effectively in 
a “hands-on” fast-response environment. 
Enormous costs of early computers, how¬ 
ever, compelled owners to convert all a- 
vailable time to productive use. Rather 
than have a multi-megabuck computer 
sit idle while a programmer scratched 
his head, jobs were assembled in batches 
and run one after another with no allow¬ 
ances for human interactions. 

The expensive components in early 
computers were the processor (CPU) 
and memory. It soon became obvious 
that many programs were being hindered 
by slow Input/Output devices. It also 
became obvious that the capacity of the 
expensive components was being wasted. 
To alleviate this problem, multiprogram¬ 
ming systems were developed which al¬ 
lowed several programs to run simultan¬ 
eously in one computer. When one pro¬ 
gram had to wait for I/O to complete, 
the processor could be used to handle 
another program. Multiprogramming 
required more memory than purely ser¬ 
ial execution, but because it used tre- 

One programmer 
put his fist through 
a keyboard after 
the computer didn’t 
respond for 
several minutes. 

mendously expensive CPU more heavi¬ 
ly and did more work in a system of 
slightly higher cost, it paid for itself. 

Multiprogramming, which originally 
shared the CPU among programs sharing 
memory, could be extended to share 
memory among programs which could 
not all fit into memory simultaneously. 
With fast disk of drum storage available, 
a program encountering a substantial de¬ 
lay could be written out and another’ 
program brought in to use the expensive 
memory and CPU. This swapping of pro¬ 
grams allowed delays longer than a sim¬ 
ple I/O wait to be accomodated without 
tying up expensive memory. 

With continued refinement of multi¬ 
programming, a solution was found for 
providing “hands-on” computing without 
dedicating an horrendously expensive 
computer to each user. If enough users 
could be collected together and connec¬ 

ted to one computer, whenever one user 
was thinking, typing-in, or receiving a 
response from the computer, that ma¬ 
chine could process work submitted by 
another user and thus keep the CPU 
busy. Because CPU speed is so great and 
almost no processor time is needed to 
type out information on the user ter¬ 
minal or receive characters from a key¬ 
board, each user generates a minimal 
amount of time over a long period. The 
timesharing user tends to request short 
bursts of computing separated by wai¬ 
ting-times measured in seconds. Hence, 
in theory, the user will not notice that 
he is sharing a computer with many peo¬ 
ple. He will have the illusion that he is 
the master of a dedicated computer, 
ready to respond to his demands. Final¬ 
ly, by inducing each user of a timeshar¬ 
ing system to pay a relatively low price 
user fee, say $20/hour, the high cost of 
the system can be recovered. 

We see, then, that the driving force 
that created timesharing was economic; 
finding a way to provide many users with 
interactive computing even though com¬ 
puters and memory were very costly. 
Once established, a timesharing utility 
can provide many useful services. Since 
many users share a common system, they 
also can share programs and data files. 

Now the same benefits are being of¬ 
fered microcomputer users with the ad¬ 
vent of microcomputer timesharing sys¬ 
tems. However, there is no need to go 
to timesharing to reap its benefits. In 
effect, we must change our attitude 
about computers, we must learn to view 
communication, not processing, as the 
key. Timesharing, as already defined, is 
the sharing of central processors and 
memory among a number of users so that 
that all users do not simultaneously re¬ 
ceive service. With the price of proces¬ 
sors and memory dropping, the practice 
that made sense in the Sixties is stupid 
in the Seventies. Timesharing is widely 
identified with benefits that evolve from 
its adoption on large computers. How¬ 
ever, because the U.S. proceeded from 
stagecoaches to trains to airplanes for 
long distance travel doesn’t mean a de¬ 
veloping country today has to go 

through the same steps. Microcompu¬ 
ter users Should pause and think before 
following the path big system people 
took into timesharing. 

What's so bad about timesharing, 
anyway? Let’s take a look at the tech¬ 
nical problems that plague timesharing. 
The most serious problem stems from 
the fact that timesharing is, after all, an 
illusion. As long as people “play by the 
rules” and interact on a line-by-line ba¬ 
sis with little computation, everything 
works fine. Before long, somebody will 
ask for a substantial amount of compu¬ 
tation, such as asking the computer to 
determine whether 2,363,383,927,257 
is prime. Because you expect this pro¬ 
cedure to take a while, you’re not 
upset when the answer doesn’t come 
back right away, and because your com¬ 
putation can “mop up” time not being 
used by other people, all is well. Suppose, 
however, that somebody else has a burn¬ 
ing desire to know the 987th digit of pi 
right at the same time you started your 
program. Now we have two mops con¬ 
tending for one CPU, and somebody is 
going to lose. If the time is equally split, 
your program will now take twice as 
long to run. And what about you? 

You’re sitting there fuming at the ter¬ 
minal wondering why what took a min¬ 
ute yesterday takes two (or five or ten 
or twenty) today. The concept of time¬ 
sharing finds its own negation in the 
following slightly cynical “Laws of 
Timesharing” (with apologies to C. 
Northcote Parkinson): 

1. Computations grow in complexi¬ 
ty to fill existing processor power. 

2. Programs grow in size to fill ex¬ 
isting memory. 

If these two widely-observed rules 
hold, the self-aggrandizement of pro¬ 
grams in a timesharing system will even¬ 
tually lead to inter-program contention 
that degrades response time. As a re¬ 

3. Timesharing is characterized by 
unpredictable variation in response time. 

The unpredictability of response time 
is the most frustrating aspect of time¬ 
sharing. Large systems can be beautiful 
for obtaining a fast answer to a big 
problem. But the system can be singu¬ 
larly infuriating when it pauses for up to 
a minute when asked to do something 
like “list my five line program” that a 
microcomputer will always do instantly. 
(Long unpredictable delays make some 
people positively homicidal. I know one 

programmer who put his fist through a 
terminal keyboard after waiting several 
minutes for a system to respond.) 

Another problem in timesharing is 
the catastrophic impact of a system 
crash or breach of security on its users. 
As systems grow more and more com¬ 
plex, there are more hardware compo¬ 
nents to break and more dark corners of 
software in which bugs may lurk. As a 
result, large timesharing systems tend to 
be less reliable than small dedicated sys¬ 
tems. Besides, they are hellishly diffi¬ 
cult to maintain and enhance. When 
the system fails, many users are simul¬ 
taneously infuriated. If the users are all 
in the same location, mobs dangerous to 
the health and welfare of systems pro¬ 
grammers can form. When many users 
share a system, they may also share pro¬ 
grams and data. Unfortunately, unless 

now in use 
make timesharing 

the system is ver-y careful, they may also 
steal or destroy each other’s programs or 
data. Millions of dollars of Government 
money have been used to develop sys¬ 
tems that do not have these problems. 
But no system exists that allows both 
cooperation and privacy. Some very 
good systems have been developed, but 
none of them has remained secure 
against pilferage attempts of clever pro¬ 

The microcomputers we now use have 
made timesharing obsolete. It is absolu¬ 
tely ludicrous to have four users sharing 
a computer that costs less than the ter¬ 
minal each one is using to access the 
system. We should take advantage of 
our position in the “third invention” of 
computers and leapfrog timesharing to 
achieve its lofty goals. If we develop 
networks to allow microcomputers to 
intercommunicate, we can then share 
programs and data without encounter¬ 
ing the infuriating problems of time¬ 
sharing. We must view communication 
as the dominant aspect of a computer’s 
function, and build software and hard¬ 
ware structures based on this concept. 

Some systems embodying this concept 
have already been built and may serve 
as a prototype for future developments. 

By proper application of networking, 
we can build systems that provide com¬ 
puters to perform most tasks rapidly, 
reliably and predictably. It should be 
a system built to the scale of a human 
being, not a corporation. Furthermore, 
it should be a system that a person can 
understand, can use effectively, and can 
maintain without managers, supervisors, 
inter-office memoranda, PERT charts, 
or a horde of narrow specialists. Such a 
system will be a network of intelligent 
components, and will intercommuni¬ 
cate with each other to allow the shar¬ 
ing and communication provided by 
current central timesharing systems. Be¬ 
cause other users will be “at arm’s 
length” over a network, it will be easier 
to build systems that are probably secure 
compared to complex centralized sys¬ 
tems where interfaces between programs 
are numerous and complex. Most user 
tasks can be done locally, on the user’s 
own machine. Only when communica¬ 
tion is required need a request be made 
that could cause unwanted delay. 

Changing economics of computer 
hardware now have solutions to prob¬ 
lems that timesharing never solved. 

Small system people like ourselves have 
the outlook, experience and indepen¬ 
dence to build our own systems. These 
special designs will return computers to 
the people they were built to help in the 
first place. It has been said that comput¬ 
er software systems tend to reflect the 
structure of the organization that pro¬ 
duces them. Is it little wonder that most 
existing systems are centralized and 
hierachical. Too many systems that 
promise “distributed processing” deliv¬ 
er, instead, a more complex terminal 
connected to an even bigger central site. 
Microcomputing, with its thousands of 
basement inventors and garage-shop en¬ 
trepreneurs, has the “organizational 
structure” to build loose, distributed, 
egalitarian, flexible and powerful sys¬ 
tems. Let’s do it. □ 

John Walker has been involved in the 
design, implementation, optimization 
and support of large-scale timesharing 
systems for over eight years . Re¬ 
cently, as a partner in Marinchip Sys¬ 
tems, he has worked on developing a 
distributed network operating system 
based on the ideas in this article. 

Reprinted by permission from Personal Computing, November 1978, ©Copyright 1978, Benwill Publishing Company.