Purpose of This Story
I know how Philo Taylor Farnsworth must have felt for much of his life.
"Philo Who?" You know, the guy that invented television. "But I
thought television was invented by RCA or Allen DuMont?" No, friends,
Farnsworth invented electronic TV. But it had to be proved in court,
and the truth was seldom acknowledged until after his death. RCA paid
royalties. Seems they agreed.
I'm damned if I want to be a Farnsworth.
(See)
I lay at least a partial claim to the genesis of the timesharing
concept, the basic reason you can read this on this Web site today. The
myth that Christopher Strachey was first with the concept has been
perpetuated in Datamation, Scientific American, and the Annals of the
History of Computing. The remarkable aspect is that Strachey himself
didn't agree! Nor does Prof. John McCarthy [6].
My claim is that I published the first paper to describe the aspects of
computer timesharing for commercial usage, with users acting individually.
It was in the public domain, published nationally. See what you think.
REFERENCES
Let's lay them up front, because they're more than references one may
wish to make secondary referral to. They're the basis for the argument!
- R.W.Bemer, "How to consider a computer", Data Control Section,
Automatic Control Magazine, 1957 Mar, 66-69
- R.W.Bemer, "The status of automatic programming for scientific computation",
Proc. 4th Annual Computer Applications Symposium, Armour Research
Foundation, 1957 Oct 24-25, 107-117 (Panel discussion, pp. 118-126).
- W.F.Bauer, "Computer design from the programmer's viewpoint",
Proc. Eastern Joint Computer Conf., Philadelphia, PA, 1958 Dec 3-5,
American Institute of Electrical Engineers, 1959 Jul, 46-51.
- C. Strachey, "Time Sharing in Large Fast Computers", Proc. Intl.
Conf. on
Information Processing, UNESCO, June 1959, B336-B341
(not on Web).
- R.M.Fano, F.J.Corbató, Time-Sharing on Computers",
Scientific American Magazine, 1966 September, 129-140.
-
See J.McCarthy, "Reminiscences on the History of Time Sharing",
1983 Winter or Spring.
- M.Campbell-Kelly, "Christopher Strachey, 1916-1975, A Biographical
Note", Annals History of Computing, 7, No. 1, 1985 Jan, 19-42.
My Unlikely Source for the Concept
I have told somewhere about attending Curtiss-Wright Technical Institute
of Aeronautics prior to World War II. The school was in Glendale, CA.
The date was the Fall of 1940. Another well-known institution there was
the Bob's Big Boy hamburger. The original! At least it made the owner
Mayor of Glendale. I went crazy for them, sometimes eating two in a row
(I was young).
The drive-in was fascinating to this Midwestern boy. Parked cars
around the restaurant, serviced by girls in attractive garb. But on
going inside you saw what was to me an absolute first -- a revolving
drum with elasticized or expanding bands. The waitresses put the orders
in there, and the cooks referred to them often. The key word is
"interspersion". Hardly any order ever got produced in its entirety at
one time. The production sequence was "shared" by all customers and by
all cooks. That idea stuck with me for 17 years!
Contribution of Robert W. Bemer (Ref. [1] - 1957 March)
The final section of this paper, commissioned by Editor Evan Herbert,
described aspects of future commercial timesharing:
"Future computer operation, which strongly influences the design of
the programming languages, has some vitally interesting possibilities.
In this glimpse, the picture presented here is dependent upon three
features:
- Faster computers always lower the dollar cost per problem
solved, but not all companies will be able to afford the
high prices of the next generation of super-computers. They
simply may not have enough problems to [fully] load one.
- Producing a spectrum of machines is a tremendous waste of
effort and money on the part of both the manufacturers and
the users.
- Availability of a huge central computer can eliminate the
discrete acquisition of multiple smaller computers, homogenize
the entire structure of usage, and allow a smaller, and more
numerous class of, user into the act, thus tapping a market
many times the size presently projected with current practice
in computer access."
"Assuming the availability of practical micro-wave communications
systems, it is conceivable that one or several computers, much larger
than anything presently contemplated, could service a multitude of
users. They would no longer rent a computer as such; instead they would
rent input-output equipment, although as far as the operation will be
concerned they would not be able to tell the difference. This peripheral
equipment would perhaps be rented at a base price plus a variable usage
charge on a non-linear basis. The topmost level of supervisory routine
would compute these charges on an actual usage basis and bill the customer
in an integrated operation.
These program features are, of course, recognizable to Operations
Research people as the Scheduling and Queuing problems. Using
commutative methods, just as motion pictures produce an image every so
often for apparent continuity, entire plant operations might be
controlled by such super-speed computers.
These future hardware capabilities (and few competent computer
manufacturers will deny the feasibility, even today, of super-speed and
interleaved programs) demonstrate a pressing need for an advanced common
language system so all users can integrate their particular operations
into the complex of control demanded by an automated future ..."
(Note: This call for a "common language" was issued two years before the
start of COBOL work.)
This excerpt shows two biases. 1) Equipment rental as a way of life,
for I was then an IBM employee, and rental was then their policy,
and 2) Manufacturing control as a primary example, for this was the
audience for the publication.
Contribution of Robert W. Bemer (Ref. [2] - 1957 October)
On p. 126, Bemer's reply was:
Q: (Omitted from the publication, but I think the word
"time-sharing" was used in the question. I used the word myself in a
talk to the Franklin Institute the previous February, but you can find
several previous papers, particularly from the SAGE project, that used
the same phrase. But strictly in a hardware sense.)
A: "I am in favor of the short-order-cook policy that I think
will come into effect perhaps five or ten years from now. It might
resolve at least a certain class of problems as between the small and
large computers. If one had an extremely large, extremely fast
centralized computer with various lines radiating out, and with terminal
facilities such as a person now only gets in in the form of input-output
devices at the computer, and if one could have high-speed transmission
to and from this centralized computer, it would be like a short-order
cook. It takes the orders off the lines and, so to speak, heats up the
griddle and sees that the toast is ready while it is pouring the coffee.
It will be self-scheduling, self-regulating, and self-billing to the
customer on the basis of use of the input-output device. I think, since
the larger and faster computers, as far as production problems are
concerned, always produce more problems solved per dollar once the
problems are in the machine, that this is the obvious direction to go.
I agree that at the present time there are many small computers that
seem to take less trouble than a large one; but I think that, in the
long run, we will use the largest computers and will start thinking in
terms of compatibility of languages and ultimately in terms of a single
language."
Contribution of Dr. Walter F. Bauer (Ref. [3] - 1958 December)
Here Walter Bauer describes his "Ultradatic, A Conjectural Computer".
The part of the paper pertinent to timesharing says:
"It is fostered by the idea that the large computer is more economical
to use today than the small computer, or a number of small computers,
as long as the total workload is sufficient to keep the large computer
busy, or as long as unnecessary expense is not incurred by idle time.
The idea further stems from the fact that with increasing frequency
one sees computers being operated from remote locations by means of
communications over voice channels, or at least extensive input or
output of data is being fed to computers from remote locations.
The central idea here is that each large metropolitan area would have
one or more of these super computers. The computers would handle a
number of problems concurrently. Organizations would have input-output
equipment installed on their own premises and would buy time
on the computer much the same way that the average household buys
power and water from utility companies. In fact, the charges for the
services would depend on the type of services rendered. If the problem
required extensive use of the higher priced circuits, for example,
high-speed multiplying, the rental rate for the time used would be
higher than in other cases. The user would be charged for that time
used and only that time used, and the accounting procedure would
reflect the customer's detailed use."
Bauer further says that "the Ultradatic idea was first described in
talks given by the author to the San Diego and Rio Grande chapters
of the Association for Computing Machinery in late 1957 and early
1958".
He refers to "parallel programming" and Stan Gill's 1957 paper on
that subject.
At the Munich IFIP meeting in August of 1962, someone mentioned my
paper to Bauer, who until that time thought he had published the first
paper on commercial timesharing. We met in a restaurant to discuss the
timings and content. Dr. Bauer, with his Ramo Wooldridge experience,
and as founder of Informatics (later bought by Sterling Software), was
usually quite up-to-date in the computer field.
Contributions of Fano and Corbató (Ref. [5] - 1966 September)
This paper goes into much detail about working TSS systems. Only the
beginning is pertinent to the purpose of this history, where the authors
credit Christopher Strachey of the United Kingdom as the one "who first
proposed (in 1959) a time-sharing system".
Contributions of McCarthy (Ref. [6])
This paper has problems of inaccuracies, suppositions, and bad dates.
McCarthy's first thinking about time-sharing "might have been around
1955". But he does say that his "first attempts to do something about
time-sharing was in the Fall of 1957 ..." And that to get the requisite
hardware from IBM "took a long time ... perhaps a year, perhaps two."
One suspects he did not keep his records.
Nor did he publish, for this paper [6] has no references. He admits
that "In all this, there wasn't much publication. I wrote a memo to
Morse dated January 1, 1959 ..." But in the next sentence he admits
that "it has been suggested that the date ... should have been 1960."
I also complained to Eric A. Weiss, of the Annals of the History of
Computing, re the genesis of timesharing. I cannot find an answer.
What I appreciate is that McCarthy complained about Strachey's credits
in [5] to the Scientific American. So did I. They replied to me
that "The point you raise is most interesting, and I have put it before
Professors Fano and Corbató".
The authors did reply to Scientific American, but oddly enough backed
off to the 1959 Jan 01 internal paper by McCarthy, 22 months after my
external paper. But McCarthy says that "Corbató was surprised
to find my 1959 memo in the files".
Also in their reply they credited Strachey with "substantial detail".
I am sure that both Bauer and I could have engendered detail to go
with the concept. I, in particular, was limited by space available in
a popular journal.
And so we come full circle. The entire chorus of credits for Strachey
hits a bad chord.
The really interesting part of McCarthy's paper is its Appendix,
containing Strachey's reply to a query from Don Knuth about origins.
Here are the significant parts of Strachey's reply:
"It was mainly about multi--programming ... I did not envisage the
sort of console system which is now so confusingly called time sharing ...
Halsbury ... was certainly right to say that in 1960 'time sharing'
as a phrase was much in the air. It was, however, generally used in
my sense rather than in John McCarthy's sense ..."
Strachey's reply also said that the IFIP Conference was in 1960. It
was in 1959. As he died just a year after writing the letter, the
memory failure is forgivable.
Contribution of Christopher Strachey (Ref. [7])
This remarkably complete story of Strachey's life confirms my tale
in many ways:
- Strachey's Stretch visit to IBM, and to the ACM Conference in
Houston in June of 1957, where we talked, three months after I had
published my timesharing paper.
- "his name is often associated with the origination of multiaccess
timesharing computers. For example, Fano and Corbató ...
credit him with the original conception." But that is second hand
reporting.
- "The paper [5] was influential and has come to be regarded by some
as a significant step in the development of time-sharing ...".
(no doubt this is true, if misguided). "Strachey was himself
surprised to receive so much credit. ... Strachey's view, somewhat
naively, was that time-sharing was in the air and that he was merely
articulating those ideas".
- Nevertheless, "Strachey filed a patent application for time-sharing
in February 1959. ... The patent (British patent 924672) was
granted in 1963." Now that's a strong articulation! The patent
seems to have had no more influence than the one recently granted
in the U.S. for the windowing method for Year 2000 repair, now
roundly laughed at.
- "A definite statement on the extent of Strachey's part in the
invention of time-sharing will have to await a more thorough analysis."
(TIME)Sharing the Credit
It has been said that great ideas often occur simultaneously to several
people. In the pertinent period there was a ferment, and one probably
cannot ascribe the entire idea to a particular person as we are able to
do with Da Vinci.
Much could have been gotten from me. As my first publication elicited
a suggestion from a Dr. Franz (???) that I should be fired because it was
not IBM's policy, that may be why IBM instigated no patent applications.
Maybe I had put it in the public domain, as I did with publication
of my escape sequence concept. But it is no secret that I was very
active professionally in the 1957-1959 period.
- Even when the New Yorker Magazine profiled me in 1957 January I
mentioned the community computer idea, which we see again
in Bauer's paper.
- I was speaking pretty much for IBM, both professionally in ACM
and to the customers personified by SHARE and GUIDE.
- My Armour Research Foundation paper became well-known,
- I knew McCarthy well, antedating the internal memorandum that
Fano and Corbató finally found, and attended his 1959 Apr 16
Symbol Manipulation conference.
- Strachey attended the ACM meeting in Houston in 1957 June,
as I did. I know, because I talked to him, as mentioned above.
- Strachey also visited IBM's STRETCH project then.
So I stand by my original claim. I did not create a timesharing
system. I did not create a practical design for such. But I was
the very first person to publish in some medium (and it was national)
the concept of sharing a computer remotely amongst many users via
input/output equipment for commercial usage!
Final Note on Historical Accuracy
As I prepare historical material for this site, I am astounded
at how many dates the historians (and those they write about)
have gotten wrong. In the fast-moving computer industry, I
would think it important to keep accurate records of idea
generation, patents, public talks and publications.
Accordingly I submit for general use the nonpersonal subset of
my travel log from 1954 (when I first kept records) to date.
I hold no copyright, and anyone wishing to flesh out the time
line may use it as a starting point. After all, busy though
I was in computer activities, I could not go to every meeting!
Each trip is identified as the nth in that year. These could be
interspersed in the text as identifiers, if anyone thinks that a
useful idea.
You can find the path to these logs in the "Interesting Computer
History" page, but I must admit that their interest is limited
to this one purpose. Otherwise, I agree. They're pretty dull!
Back to History Index
Back to Home Page