The following articles are reprinted solely as items of interest for the independent evaluation by members of
ATSU. The opinions, statements of fact, and conclusions expressed herein are not those of the Association.
a perspective on the IBM 5100
Some crackerbarrel philosophizing on the historical and
competitive significance of this unique package.
The IBM 5100 Portable Computer represents a unique devel¬
opment in computer history, but its uniqueness rests nei¬
ther on its size, price or portability. Even the Hewlett-
Packard Model 65 programmable scientific pocket calcula¬
tor, introduced in January 1974 at under $800, excels the
5100 in each of those three criteria. Neither is the 5100
unique in providing alphanumeric I/O and auxiliary mass
memory in a portable, interpretive-language computer.
Wang Computer Laboratories, among others, has been offer¬
ing such systems for about a year, and at some two-thirds
the price.
What is unique about the 5100 is the vast amount of
memory it contains. Memory capacity is still the largest
single cost element in a minicomputer, and, particularly in a
computational system, by far the most important single
factor in rating its power. The reason for this is that unlike,
for example, bandwidth in a communication system, mem¬
ory capacity is rarely negotiable. One can usually trade off
the cost of more connect time against the cost of wider-
band (i.e., faster) lines, but there are many classes of prob¬
lems that cannot even be attempted at memory capacities
below, say, 16K bytes. There is simply nothing to trade.
Speed, the most important cost factor in choice of mem¬
ory, is, of course, also related to performance. However, as
opposed to minis for real-time applications, where even a
single millisecond can be critical, extra seconds are rarely
noticed in small-scale computational systems.
PROGRAMMABLE SYSTEMS
Programmability and memory are interdependent, but
not identical. A microprocessor can handle simple algo¬
rithms quite well, and it doesn’t care if the data comes from
internal semiconductor storage, external magnetic storage,
or is entered digit-by-digit from a keyboard. The difficulty
comes when the needed algorithm is not preprogrammed.
Using those that are to solve those that aren’t is what pro¬
gramming is all about. But a “closed” system limited to
only a hundred or so very slow-acting arithmetic statements
is a computer in name only. It is hardly what a software
specialist would have in mind when describing a “program¬
mable system.” That label properly applies only to systems
capable of handling a full repertoire of branching, looping,
nesting, etc. operations. The fact that a low-priced system
might still not be able to do very much with those capabili-
MOST VERSATILE. The Hewlett-Packard Series 9800, Model 30
calculator, shown here with page-width thermal printer, is offered
with a broader line of peripherals and special-purpose interfaces
than any other desktop unit.
COMPUTER AT A CALCULATOR PRICE. The Wang System
2200-S, a full-function BASIC system from the most experienced
supplier of desktop interactive computers.
ties (i.e., memory capacity might still be low) becomes less
important than that the capabilities at least exist. “Pro¬
grammability,” or the ability to manipulate data, precedes
“program ability,” or the ability to describe entire
procedures.
The first approach to reducing the cost of programmable
systems to the point where they could be economically jus¬
tified by individual researchers was time-sharing. While in-
house time-sharing systems with per-terminal costs below
$10,000 were available prior to 1970, their overall cost
typically ran well over $250,000. Small users had nothing
to compare with the offerings of commercial T-S services.
By 1970, most of the larger T-S services were offering
BASIC as their standard, interactive computational lan¬
guage, and in-house systems manufacturers followed suit.
One of the first companies to offer an in-house, BASIC-
language T-S system around that time was Hewlett-Packard.
In January 1970, a feature article in MODERN DATA
described the System I from Interplex Corp. of Waltham,
MA, to our knowledge the first in-house system with dedi¬
cated BASIC-language terminals, albeit with numeric-only
output. Using a Honeywell H-316 processor, the Interplex I
could handle up to 16 such terminals, and only five were
necessary to bring the per-terminal cost to under $10,000.
Later that same year, Wang Laboratories introduced the
3300, also a 16-user BASIC system, but built around a pro¬
prietary byte-oriented TTL processor and much lower in
price. (Interplex subsequently folded its tent.)
These early low-cost T-S systems became quite popular,
although the amount of memory available to each user was
invariably under 16K bytes. Today there are several 16-
and 32-user in-house T-S systems that provide considerably
more memory to each user, and at a per-terminal price well
under $10,000. But by the end of 1973, sophisticated
pocket calculators had become very popular, and attention
shifted to stand-alone systems with internal processors.
Quite logically, people wondered how much more could be
done with a box that was not constrained to fit in a pocket.
They had not long to wait.
BASIC ON A DESK
“Programmable” desk-top calculators had^ of course,
been around since the late Sixties (e.g., Wright Line’s “Math-
atron” and Wang Laboratories’ “LOCI”), but they were
hardly a match even for today’s numeric-only pocket ver-
sions.The first two breakthroughs were the Hewlett-Packard
9830A and the Wang Laboratories 2200-S, both introduced
in 1974, both BASIC-language systems and both available
with a full range of peripherals.
The 9830A is conservatively described as a desk-top cal¬
culator. Actually, it is a full-function (e.g., multi-level nest¬
ing, conditional transfers, etc.) BASIC-language computer,
consisting of 16K bytes of ROM, up to 16K bytes of RAM,
32-character LED display, typewriter keyboard and 64K
byte cassette drive. A 9830A with 16K bytes of user-
available RAM is $10,560. Although priced higher than the
5100, the 9830A is considerably more versatile. Options
include eleven plug-in ROM modules - including three for
data communications - eight I/O peripherals, special inter¬
face cards for test instrument and control applications, and
more than fifty software packages in eight application areas.
The Wang Laboratories 2200-S, $lso a full-range BASIC
computer, was introduced at the end of 1974. It is a two-
box system, consisting of a separately-packaged processor/
memory unit and a console with 9”, 1024-character CRT,
cassette drive and keyboard. Total weight of the two cab¬
inets is less than 90 lbs
The 2200-S comes in at $5,400 with 24K bytes of ROM
and 4K bytes of RAM, and tops out at $12,600 with 32K
bytes of RAM. At the 16K bytes level, the 2200-S is less
expensive than the 16K byte, BASIC-only model of the
5100: $8,600 vs. $8,975. Indeed, this $375 difference, ac¬
cording to Wang Labs’ Edward Lesnick, is only the tip of
the iceberg. The 2200-S is available in a 12K bytes configur¬
ation, of which only 700 bytes are reserved from the user,
for $7,400. Lesnick maintains that 4,400 bytes of the
5100’s RAM are always unavailable for user storage, there¬
by giving Wang a $1,575 price advantage. Lesnick also
claims that the 5100 is slower and less efficient than the
2200-S in its use of memory. As for the top end of the
5100 line, Lesnick points to the WCS-20, an all-in-one-desk
version of the 2200-S with 42K bytes of ROM, 8K bytes of
RAM, 12” CRT, keyboard and 250K bytes of floppy disk
for $11,000. Overlays called in from the floppy would do
for “most” applications, says Lesnick, and for another
$3,000, two additional floppies could be added.
Lesnick’s argument’s regarding 5100 competition at the
16K level are strong, as is Wang Labs’ software and peri¬
pherals support for the 2200-S. But where the WCS-20 is
indeed better suited to many applications than upper-level
5100’s, it is precisely because these two systems are not
comparable. RAM, and lots of it, is what the 5100 is all
about. Overlays and auxiliary storage are not.
THREE FROM DEC
Consider, for example, Digital Equipment’s “triple
announcement” earlier this year: the commercial Data-
system 310; the CMS/1 interactive engineering system; and
the “Classic,” an interactive system for educational applica¬
tions - all centered around PDP-8/A minis. Only one of
these systems is likely to be affected by the 5100.
DEC’s “Classic” gives the user access to 20K 12-bit
words (of the 32K provided) of core RAM; a VT50, 12 x
80-character CRT with electrolytic copier; dual floppies;
and BASIC. RAM is certainly adequate for everyday educa¬
tional use, and the floppies provide a large measure of flex¬
ibility. The “Classic” sells for $7,900. A 16K BASIC 5100
with printer is $12,650.
The Datasystem 310 is a disk-based business system for
under $15,000. If you need a disk-based business system,
you should have a disk-based business system; not a 5100.
The CMS/1 offers a nominal 32K of RAM with a VT50
and a set of dual floppies for $12,000, and makes the
electrolytic printer and a 30 cps DECwriter available as
options. For software, Fortran IV is standard; BASIC and
COGO, the civil engineering language, are options. This one
could be affected.
ALL A MATTER OF MEMORY
In any case, like the Hewlett-Packard and Wang systems
mentioned previously, the CMS/1 is not, nor was it de¬
signed, for programs that require upwards of 32K bytes of
RAM. If you need that much memory in a small thinking
machine, the 5100 is the only way to go. If you don’t need
that much, and peripheral flexibility is important, there are
many ways to go. ■
Reprinted with permission from MODERN DATA, October 1975.
NEW PRODUCTS
More flexibility for data banks
New ‘relational’ systems
simplify storage, answer
impromptu questions
"I’m ecstatic over it,” exclaims John F.
Bartol, a medical data analyst at
Hoechst-Roussel Pharmaceuticals Inc.
of Somerville, N. J., a subsidiary of
American Hoechst Corp. Bartol is talk¬
ing about a new computerized manage¬
ment information system called No¬
mad. He uses it to prepare reports on
clinical drug tests required by the Food
& Drug Administration. "Nomad en¬
ables us to respond almost immediately
to impromptu questions,” he explains.
If so, says an outside software expert,
"that would represent a tremendous
breakthrough.”
Nomad was announced late last
month by National CSS, a Norwalk
(Conn.) company that sells computer
time-sharing services. At the time
NCSS’s product manager, James
McGuire, boasted that it "will give us a
competitive edge” over the other com¬
panies in the field. But the edge lasted
about two weeks-until Tymshare Inc.
in Cupertino, Calif., announced a simi¬
lar system called Magnum. McGuire in¬
sists, however, that Nomad is easier to
use because it responds to simpler Eng¬
lish commands. Tymshare’s product
manager, Thomas Tranfaglia, takes is¬
sue with this claim, of course.
Both NCSS and Tymshare are rela¬
tively small companies in the $500 mil¬
lion a year time-sharing industry-with
NCSS revenues running at $33 million
annually and Tymshare’s at $46 mil¬
lion. But the two publicly owned com¬
panies are also the largest indepen¬
dents among some 58 contenders that
include General Electric Co/s Informa¬
tion Systems Div. and Control Data
Corp.’s Service Bureau Corp.
Time sharing permits many users to
access a computer from different re¬
mote locations by means of a telephone
and computer terminal. It was first
commercialized as a calculating tool for
scientists and engineers, but now man¬
agement information, or data base,
systems are the fastest growing part
of the time-sharing business.
Nomad and Magnum are both aimed
at the same kind of job-typically to
convert voluminous raw data on sales,
inventories, production scheduling, test
information, and the like into sum¬
mary reports. Other management in¬
formation systems do that too, but the
two new ones are the first to put into
commercial use a more efficient "rela¬
tional” technology for organizing the
data.
Conventional data bases use a "hier¬
archical” structure that is limited to
filing and retrieving all the data-base
items through a link to one key refer¬
ence. Relational systems, developed
theoretically in the late 1960s, over¬
come this inflexibility by linking the
data with cross references.
For example, in a hierarchical data
base for a parts inventory system, the
suppliers probably will be the key ref¬
erence. Even if the user wants data
such as part numbers, units on hand, or
slipped deliveries, he must program his
requests by means of the suppliers’
identities. A relational system does not
organize data in this rigid fashion, and
field-test users report many benefits:
simpler storage: Ethyl Corp/s engi¬
neering department in Baton Rouge,
La., uses Tymshare’s Magnum to keep
track of the time that some 40 drafts¬
men spend on various projects. Under
the former hierarchical system, ex¬
plains design supervisor Samuel
Brown, the data for separate reports,
such as overtime and total hours
worked on each project, had to be kept
in separate computer files. "Otherwise,
we could not break out the two sets of
data independently,” he says. "With
Magnum, we now use only one data
bank and can turn out any kind of re¬
port we want in one pass.”
MORE FLEXIBLE FORMAT: Relational sys¬
tems enable a user to select a printout
with any arrangement of columns,
headings, margins, totals, and so on.
With hierarchical systems, the format
is fixed. That is why NCSS’s Nomad ap¬
peals to Richard J. Hannon, manager
of marketing information services at
U. S. Pharmaceutical Co., an E. R.
Squibb & Sons subsidiary. Hannon
keeps computerized sales records for 12
product managers on 1,750 items.
"A good-looking report is critical in
marketing,” he says, "and Nomad has
eliminated about all the deficiencies in
formating that I have ever seen in
other data-base systems.” For ex¬
ample, columns may be titled with
words of a user’s choice rather than
with code letters, and there are both a
zero symbol and a "not-available” ab¬
breviation instead of the double-duty,
ambiguous zero.
Donald E. Seese, controller at Wells
Fargo Bank in San Francisco, concurs.
He is converting the bank’s profit plan¬
ning and functional cost file to Mag¬
num, partly because "we will be able to
make formating changes that we
couldn’t do previously.” Says NCSS’s
McGuire: "It is unique that a general¬
ized program can do specialized for¬
mating.”
VARIED OUTPUT: Relational systems
can turn out responses to unforseen
questions, unlike hierarchical systems
where all reporting needs must be de¬
termined when they are set up. New
England Merchants National Bank in
Boston, for example, is putting "every
bit of information” on its certificates
of deposit into a Nomad data base, says
a spokesman. "We don’t even need to
anticipate the usefulness of the data. If
anything unforseen occurs, the ma¬
chine has stored the information
needed to prepare a report.”
Users also point to big operational
savings. Ethyl’s Brown was spending
$1,500 to $2,000 per month simply to
store data in his old system. "Magnum
should be able to cut that storage cost
by 50%,” he says. And he anticipates
saving 15% to 20% more in processing
time. Bartol of Hoechst-Roussel like¬
wise claims big savings. "It costs 850 to
develop and venerate an output
request with Nomad,” he says. "Under
our previous system, it cost $4.50.”
Some users look more for added op¬
portunities than for potential savings.
Squibb’s Hannon says he expects no re¬
duction in his department’s budget be¬
cause, "I find that I am now able to use
our Nomad data base in so many more
ways than I would under our prior sys¬
tem.” The higher unit prices that both
companies are charging for their new
systems may also limit savings.
Some industry people do not expect
the new system to catch on widely be¬
cause most big companies already have
data bases in place. To convert them
would take huge investments in money
and retraining, and there would be
chaos during the transition, they say.
Even McGuire concedes that Nomad
will be used primarily in setting up
new data bases. But, he adds, this still
provides a market that is growing at a
rate of 25% a year. "For new data
bases,” he says, "relational technology
is the way the industry will go.” ■
Reprinted with permission from BUSINESS WEEK), October 27,. 1975.
Time On Their Hands
Once there were 200 time-sharing companies.
Now there are only 50. Soon there will be...?
George J. Feeney runs General Elec¬
tric's Information Services division.
He is also the most eloquent prophet
of computer time-sharing—the busi¬
ness where customers rent time on
someone else’s computers instead of
tying up capital to buy their own.
“Computing will increasingly be
purchased as a utility service much
like electric power or water,” Feeney
has prophesied. “The major vendors
will truly resemble utilities. They will
provide computer power to virtually
every community seven days a week,
24 hours a day, with services to
match the needs of small businesses,
research and development organiza¬
tions and worldwide corporations.”
If George Feeney is right GE, Con¬
trol Data, Computer Sciences Corp.
and several others currently building
up huge international time-sharing
networks will be in clover. For now,
however, the time-sharing business is
more like a nightmare. A red-hot con¬
cept just ten years ago, with revenues
growing by 25% a year, time-sharing
was attracting new entrants at the
rate of nearly one a week in the late
Sixties. At one point some 200 com¬
panies offered customers simultaneous
access to master computers from ter¬
minals located in their own offices or
plants. Today only 50 companies re¬
main, and by the end of the Sev¬
enties, by wide agreement, the num¬
ber may well be down to ten or fewer.
Present survivors, like Tymshare,
National CSS and Automatic Data Pro¬
cessing (see table), have tried to im¬
prove their staying power by buying
up the businesses of some of the strag¬
glers. “Of die 50-odd companies still
around,” says President Robert J.
O'Brien of Rapidata, “20% are losing
money, 10% are barely breaking even
and the rest are making anywhere
from 5% to 20% on sales after taxes.”
What happened to a business that
looked so promising just a few years
ago? Overcrowding is an obvious an¬
swer. But if the business had mea¬
sured up to the glowing expectations,
there would have been plenty of
room for almost everyone. Predic¬
tions were that the business would be
running $1.8 billion a year by 1975;
actual revenues are estimated by a
recent Quantum Science survey to be
half that.
The Little Devils
The minicomputer revolution, which
also got rolling in the late Sixties,
caused the first nightmares for time¬
sharing boys. Back when it was sim¬
ply a choice between renting time on
someone else's equipment and invest¬
ing in a large or even standard-sized
computer of one's own, customers
came a-running to the time-sharing
outfits. But when it became cheaper
for a small businessman or the re¬
search division of a large corporation
to buy a mini rather than rent time,
the squeeze was on in time-sharing.
More recently, Sperry Rand, Interna¬
tional Business Machines and other
mainframe manufacturers have begun
selling new lines of communications-
linked computers that let customers
set up their own networks. Another
setback for time-sharing.
George Feeney's response is the
idea of computer utilities with global^
round-the-clock time-sharing net¬
works. GE's Mark III network, in
which it has invested some $200 mil¬
lion so far, stretches all the way from
western Europe to the western Pa¬
cific. The others, for the most part,
cover only sections of the U.S. and
Europe. Tied in with these networks
are data-storage banks and broadly
designed software programs that are
intended both to make the networks
more useful to customers and to lock
the customers more tightly into them.
Using a Computer Sciences setup,
for example, all U.S. Army recruiters
can instantly place a new recruit's
special skills and preferences into a
data bank, schedule him for basic
training and enroll him in special
schools. No one is sure what it would
cost to duplicate such a system, but
it would be extremely expensive; its
monthly rental fee to CSC is $140,000.
Telephone companies are fairly big
users of time-sharing for scheduling
work forces, keeping track of repair
records and designing entire telephone
systems. They account for perhaps
10% of all revenues. The biggest cus¬
tomer by far, however, is the U.S.
Government. This year Uncle Sam
will spend about $240 million for
various time-sharing services, twice
last year's outlay.
Even if the overall growth proves a
bit disappointing in time-sharing,
there will be demand from big com¬
panies that need extra computer time
during peak-load periods. The busi¬
ness is here to stay.
Under the circumstances, it is hard
to see how General Electric can lose.
It has the money and the reputation,
and it got an early start. But the fu¬
ture for many of the smaller partici¬
pants is dim: In a business like this,
foresight isn't enough; you need stay¬
ing power, too. ■
The Survivors /
These ten companies, along with the captive operations of large companies like GE,
Control Data and Boeing, are among the likely survivors in the time-sharing shakeout.
Company
latest
Annual
Revenues
(millions)
Change
From
Prior
Year
Latest
Annual
Earnings
Per Share
Change
From
Prior
Year
Return
On
Capital
Recent
Price
—-Stock Data--
1975
Price Range
p/e'
Ratio
Automatic Data Processing
$154P
26%
$1.98P
19%
18.3%
52
65 -27<A
29
Computer Network
3.8
-1
-0.46
P-D
def
2%
2%- 'k
—
Computer Sciences
177.4
21
0.26
116
7.8
4 Vi
6%- 1 3 A
19
Corn-Share
12.3
29
0.55
-8
10.8
2%
4 Vs- 2%
4
Keydata Corp
11.7
23
-0.22
P-D
def
2%
3%- 1%
-
National CSS
32.3
37
1.67
16
19.9
10
14 - 5 3 A
6
On-Line Systems
11.4
15
1.50
-22
15.6
13%
17%- 8
10
Rapidata
11.5
21
0.28
-30
11.3
3Vi
5’A- Vh
8
Tymshare
46.5
32
0.85
55
19.0
17 3 A
20% - 6 3 A
18
Wyiy Corp
80.4
-3
-1.28
D-D
def
3
4%- 1%
—
NA—Not Available. D~D—Deficit to Deficit. P-D—Profit to Deficit.
P—Preliminary.
def.-Deficit.
Reprinted with permission from FORBES, September 1, 1975.
