This is a talk I gave at Papers We Love NYC on April 12, 2017. The talk is based on some papers about this history of COBOL and how it relates to computer science.

1. COBOL and Computer
Science
Andrew Turley (@casio_juarez)
Papers We Love NYC
April 12, 2017

2. The paper ...
“The Relationship Between COBOL and
Computer Science”
Ben Shneiderman
Annals of the History of Computing ( Volume: 7, Issue: 4, Oct.-Dec. 1985 )
But also: “The Early History of COBOL” by Jean Sammet

3. The Relationship Between COBOL And Comp Sci
Relationship Status ...
Computer
Science
COBOL

4. The Relationship Between COBOL And Comp Sci
Shneiderman presents three perspectives
● Historical
● Technical
● Social/Psychological

5. History

6. History -- April 1959
At a meeting of “computer people” at the University of Pennsylvania Computer
Center there is a discussion of the need for a “machine independent common
language”

7. History -- May 1959
Meeting held at the Pentagon to discuss the need for a common business
language
About 40 participants:
● 15 from government
● 11 users and consultants
● 15 representatives of manufacturers

8. History -- May 1959 (continued)
Three committees were created to develop a programming language (executive,
short-range, long-range) with members from …
● David Taylor Model Basin
● Sylvania Electric Products
● Sperry Rand
● US Steel
● DuPont
● Air Materiel Command
● NCR
● USA - Signal Corps ADPS School
● National Bureau of Standards
● Department of the Navy
● IBM
● US Air Force
● Honeywell
● RCA
● Burroughs

9. History -- June - December 1959
The short-range committee explores language options “to recommend a short-
range composite approach (good for at least the next year or two)”

10. History -- June - December 1959
“It was definitely felt that the Intermediate-Range Committee would have the time
and resources to develop a really good business data processing language.”
-- Sammet

11. History -- June - December 1959
“It was definitely felt that the Intermediate-Range Committee would have the time
and resources to develop a really good business data processing language.”
-- Sammet
Narrator Voice: They didn’t.

12. History -- June - December 1959
“I am certainly convinced in my own mind that had the Short-Range Committee
realized at the outset that the language it created was going to be in use for such
a long period of time, it would have gone about the task quite differently.”
-- Sammet

13. History -- December 1960
Same program runs on two different computers.
First real cross-platform language!

14. History -- August 1961
“COBOL: A Sample Problem” appears in Communications of the ACM
7 pages long:
● 2.5 pages of flow charts
● 3 pages of code
● 0 references

15. History -- May 1962
Communications of the ACM issue dedicated to COBOL
● 13 articles about COBOL
● Only 4 have references

16. Technology

17. Technology
Goals:

18. Technology
Goals:
● Solve business problems
EMPLOYEES.DAT
JOHN SMITH 13
SALLY JONES 42
ROCK MANLY 42
...
PROJECTS.DAT
DEATH-RAY 13
SHARK-BOMB 42
KITTEN-TOES 78
...
PROJECT-HEADCOUNT.DAT
DEATH-RAY 01
SHARK-BOMB 02
KITTEN-TOES 00
...
CREATE-
PROJECT-
HEADCOUNT.
CBL

19. Technology
Goals:
● Solve business problems
● Cross-platform (“common”)
PROGRAM.CBL
OUTPUT.DAT

20. Technology
Goals:
● Solve business problems
● Cross-platform (“common”)
● English-like language
PERFORM REPORT-GENERATION UNTIL END-SAMPLE-DATA.

21. Technology
Goals:
● Solve business problems
● Cross-platform (“common”)
● English-like language
● Verb-base language, “few
verbs with many options
rather than a large number
of verbs with a few options”

22. Technology
COBOL program
● Divisions (IDENTIFICATION, ENVIRONMENT, DATA, PROCEDURE)
○ Sections
■ Paragraphs
● Sentences
○ Statements

23. Technology
IDENTIFICATION DIVISION.
* program name
ENVIRONMENT DIVISION.
* environment-dependent stuff
DATA DIVISION.
* define record (data) structure
PROCEDURE DIVISION.
* define operations

24. Technology
IDENTIFICATION DIVISION.
* program name
ENVIRONMENT DIVISION.
* environment-dependent stuff
DATA DIVISION.
* define record (data) structure
PROCEDURE DIVISION.
* define operations
IDENTIFICATION DIVISION.
PROGRAM-ID.
SAMPLE.
AUTHOR.
CASIO JUAREZ.
DATE-WRITTEN.
04/06/17.

25. Technology
IDENTIFICATION DIVISION.
* program name
ENVIRONMENT DIVISION.
* environment-dependent stuff
DATA DIVISION.
* define record (data) structure
PROCEDURE DIVISION.
* define operations
ENVIRONMENT DIVISION.
INPUT-OUTPUT SECTION.
FILE-CONTROL.
SELECT SAMPLE-FILE ASSIGN TO "S.DAT"
ORGANIZATION IS LINE SEQUENTIAL.

26. Technology
IDENTIFICATION DIVISION.
* program name
ENVIRONMENT DIVISION.
* environment-dependent stuff
DATA DIVISION.
* define record (data) structure
PROCEDURE DIVISION.
* define operations
DATA DIVISION.
FILE SECTION.
FD SAMPLE-FILE.
01 SAMPLE-DATA.
88 END-SAMPLE-DATA VALUE HIGH-VALUES.
02 SAMPLE-NAME.
03 FIRST-NAME PIC X(8).
03 LAST-NAME PIC X(8).
02 DATE-OF-BIRTH.
03 YOB PIC 9(4).
03 MOB PIC 9(2).
03 DOB PIC 9(2).
WORKING-STORAGE SECTION.
01 GENERATION PIC X(10).
01 BOOMER PIC X(10) VALUE "boomer".
01 GEN-X PIC X(10) VALUE "gen X".
01 MILLENNIAL PIC X(10) VALUE "millennial".
01 UNKNOWN PIC X(10) VALUE "???".
* define record (data) structure
Input Record Format (SAMPLE-DATA):
|_ _ _ _ _ _ _ _|_ _ _ _ _ _ _|_ _ _ _|_ _|_ _|
| FIRST-NAME | LAST-NAME | YOB |MOB|DOB|
| SAMPLE-NAME | DATE-OF-BIRTH |
| SAMPLE-DATA |

27. Technology
IDENTIFICATION DIVISION.
* program name
ENVIRONMENT DIVISION.
* environment-dependent stuff
DATA DIVISION.
* define record (data) structure
PROCEDURE DIVISION.
* define operations
PROCEDURE DIVISION.
BEGIN.
OPEN INPUT SAMPLE-FILE.
READ SAMPLE-FILE
AT END SET END-SAMPLE-DATA TO TRUE.
PERFORM REPORT-GENERATION UNTIL END-SAMPLE-DATA.
CLOSE SAMPLE-FILE.
STOP RUN.
REPORT-GENERATION.
IF YOB >= 1980
SET GENERATION TO MILLENNIAL
ELSE IF YOB >= 1946 AND YOB < 1965
SET GENERATION TO BOOMER
ELSE IF YOB >= 1965 AND YOB < 1980
SET GENERATION TO GEN-X
ELSE
SET GENERATION TO UNKNOWN.
DISPLAY FIRST-NAME SPACE LAST-NAME SPACE GENERATION.
READ SAMPLE-FILE
AT END SET END-SAMPLE-DATA TO TRUE.
Input Record Format (SAMPLE-DATA):
|_ _ _ _ _ _ _ _|_ _ _ _ _ _ _|_ _ _ _|_ _|_ _|
| FIRST-NAME | LAST-NAME | YOB |MOB|DOB|
John Smith 19500210
Stacy Jones 19781015
Kendra Manning 19840323

28. Technology
IDENTIFICATION DIVISION.
* program name
ENVIRONMENT DIVISION.
* machine-dependent stuff
DATA DIVISION.
* define record (data) structure
PROCEDURE DIVISION.
* define operations
PROCEDURE DIVISION.
BEGIN.
OPEN INPUT SAMPLE-FILE.
READ SAMPLE-FILE
AT END SET END-SAMPLE-DATA TO TRUE.
PERFORM REPORT-GENERATION UNTIL END-SAMPLE-DATA.
CLOSE SAMPLE-FILE.
STOP RUN.
REPORT-GENERATION.
IF YOB >= 1980
SET GENERATION TO MILLENNIAL
ELSE IF YOB >= 1946 AND YOB < 1965
SET GENERATION TO BOOMER
ELSE IF YOB >= 1965 AND YOB < 1980
SET GENERATION TO GEN-X
ELSE
SET GENERATION TO UNKNOWN.
DISPLAY FIRST-NAME SPACE LAST-NAME SPACE GENERATION.
READ SAMPLE-FILE
AT END SET END-SAMPLE-DATA TO TRUE.
Paragraphs
Significant
whitespace

29. Technology
IDENTIFICATION DIVISION.
* program name
ENVIRONMENT DIVISION.
* machine-dependent stuff
DATA DIVISION.
* define record (data) structure
PROCEDURE DIVISION.
* define operations
PROCEDURE DIVISION.
BEGIN.
OPEN INPUT SAMPLE-FILE.
READ SAMPLE-FILE
AT END SET END-SAMPLE-DATA TO TRUE.
PERFORM REPORT-GENERATION UNTIL END-SAMPLE-DATA.
CLOSE SAMPLE-FILE.
STOP RUN.
REPORT-GENERATION.
IF YOB >= 1980
SET GENERATION TO MILLENNIAL
ELSE IF YOB >= 1946 AND YOB < 1965
SET GENERATION TO BOOMER
ELSE IF YOB >= 1965 AND YOB < 1980
SET GENERATION TO GEN-X
ELSE
SET GENERATION TO UNKNOWN.
DISPLAY FIRST-NAME SPACE LAST-NAME SPACE GENERATION.
READ SAMPLE-FILE
AT END SET END-SAMPLE-DATA TO TRUE.
PERFORM

30. Technology
IDENTIFICATION DIVISION.
* program name
ENVIRONMENT DIVISION.
* machine-dependent stuff
DATA DIVISION.
* define record (data) structure
PROCEDURE DIVISION.
* define operations
PROCEDURE DIVISION.
BEGIN.
OPEN INPUT SAMPLE-FILE.
READ SAMPLE-FILE
AT END SET END-SAMPLE-DATA TO TRUE.
PERFORM REPORT-GENERATION UNTIL END-SAMPLE-DATA.
CLOSE SAMPLE-FILE.
STOP RUN.
REPORT-GENERATION.
IF YOB >= 1980
SET GENERATION TO MILLENNIAL
ELSE IF YOB >= 1946 AND YOB < 1965
SET GENERATION TO BOOMER
ELSE IF YOB >= 1965 AND YOB < 1980
SET GENERATION TO GEN-X
ELSE
SET GENERATION TO UNKNOWN.
DISPLAY FIRST-NAME SPACE LAST-NAME SPACE GENERATION.
READ SAMPLE-FILE
AT END SET END-SAMPLE-DATA TO TRUE.

31. Technology
At first there was only one string processing command (INSPECT) and it was
pretty weak

32. Technology
Every sentence could end with “AND GO TO …”
🍝

33. Technology
No local variables, no real procedure calls

34. Psychology

35. Psychology
Data processing was not seen as an interesting problem

36. Psychology
Professors were not interested in running “trade schools”

37. Psychology
English seemed “unscientific”

38. Conclusions

39. Conclusions
COBOL ignored academic computer science (for better or worse)

40. Conclusions
COBOL was widely adopted and used

41. Conclusions
COBOL was a flawed language

42. Thanks!
Further Reading
● “The Relationship Between COBOL and Computer Science” Ben
Shneiderman
● “The Early History of COBOL” Jean Sammet
● “Learning To Program In Structured COBOL: Parts 1 & 2” Yourdon et al
● Communications of the ACM, May 1962
● Micro Focus (microfocus.com) -- Go to a COBOL Dev Day!

45. A Brief Introduction to COBOL

46. A Brief Introduction to COBOL
COmmon → write once, run anywhere!
Business Oriented → record processing, extract-transform-load (ETL)
Language → English-like syntax (verbs, nouns, sentences, paragraphs)

47. A Brief Introduction to COBOL -- Code
IDENTIFICATION DIVISION.
* program name
ENVIRONMENT DIVISION.
* machine-dependent stuff
DATA DIVISION.
* define record (data) structure
PROCEDURE DIVISION.
* define operations

48. A Brief Introduction to COBOL -- Code
IDENTIFICATION DIVISION.
* program name
ENVIRONMENT DIVISION.
* machine-dependent stuff
DATA DIVISION.
* define record (data) structure
PROCEDURE DIVISION.
* define operations
IDENTIFICATION DIVISION.
PROGRAM-ID. SAMPLE.

49. A Brief Introduction to COBOL -- Code
IDENTIFICATION DIVISION.
* program name
ENVIRONMENT DIVISION.
* machine-dependent stuff
DATA DIVISION.
* define record (data) structure
PROCEDURE DIVISION.
* define operations
ENVIRONMENT DIVISION.
INPUT-OUTPUT SECTION.
FILE-CONTROL.
SELECT SAMPLE-FILE ASSIGN TO "S.DAT"
ORGANIZATION IS LINE SEQUENTIAL.

50. A Brief Introduction to COBOL -- Code
IDENTIFICATION DIVISION.
* program name
ENVIRONMENT DIVISION.
* machine-dependent stuff
DATA DIVISION.
* define record (data) structure
PROCEDURE DIVISION.
* define operations
DATA DIVISION.
FILE SECTION.
FD SAMPLE-FILE.
01 SAMPLE-DATA.
88 END-SAMPLE-DATA VALUE HIGH-VALUES.
02 SAMPLE-NAME.
03 FIRST-NAME PIC X(8).
03 LAST-NAME PIC X(8).
02 DATE-OF-BIRTH.
03 YOB PIC 9(4).
03 MOB PIC 9(2).
03 DOB PIC 9(2).
WORKING-STORAGE SECTION.
01 GENERATION PIC X(10).
01 BOOMER PIC X(10) VALUE "boomer".
01 GEN-X PIC X(10) VALUE "gen X".
01 MILLENNIAL PIC X(10) VALUE "millennial".
01 UNKNOWN PIC X(10) VALUE "???".
* define record (data) structure
Input Record Format (SAMPLE-DATA):
|_ _ _ _ _ _ _ _|_ _ _ _ _ _ _|_ _ _ _|_ _|_ _|
| FIRST-NAME | LAST-NAME | YOB |MOB|DOB|
| SAMPLE-NAME | DATE-OF-BIRTH |
| SAMPLE-DATA |

51. A Brief Introduction to COBOL -- Code
IDENTIFICATION DIVISION.
* program name
ENVIRONMENT DIVISION.
* machine-dependent stuff
DATA DIVISION.
* define record (data) structure
PROCEDURE DIVISION.
* define operations
PROCEDURE DIVISION.
BEGIN.
OPEN INPUT SAMPLE-FILE.
READ SAMPLE-FILE
AT END SET END-SAMPLE-DATA TO TRUE.
PERFORM REPORT-GENERATION UNTIL END-SAMPLE-DATA.
CLOSE SAMPLE-FILE.
STOP RUN.
REPORT-GENERATION.
SET GENERATION TO UNKNOWN.
IF YOB >= 1980
SET GENERATION TO MILLENNIAL
ELSE
IF YOB >= 1946 AND YOB < 1965
SET GENERATION TO BOOMER
ELSE
IF YOB >= 1965 AND YOB < 1980
SET GENERATION TO GEN-X.
DISPLAY FIRST-NAME SPACE LAST-NAME SPACE GENERATION.
READ SAMPLE-FILE
AT END SET END-SAMPLE-DATA TO TRUE.

52. A Brief Introduction to COBOL -- Historical Position
Steve Russell writes the first implementation of LISP → (early? late?) 1959
Discussion of developing “a problem-oriented but machine independent common
language for business problems” → UPenn April, 1959
“Recursive Functions of Symbolic Expressions And Their Computation By
Machine” → CACM April, 1960
“Essentially the same COBOL program was run on [two computers]” → December
1960
“COBOL: a sample problem” → CACM August, 1961
The COBOL issue → CACM May, 1962

53. A Brief Introduction to COBOL -- Assumptions
Technical
● business users want to read
records from files, process them,
and write new records to other
files
● records have a fixed number of
fixed-size fields
● files are generally stored on
separate storage media (like
tapes)
Non-Technical
● “managers” (non-technical
people) should be able to read
and write code
● making a programming language
similar to a “real” language will
make it easier to read and write
● mathematical notation is off-
putting

54. “The Relationship Between COBOL and
Computer Science”
Ben Shneiderman, 1985

55. The Relationship Between COBOL and Comp Sci
“For a computer scientist to write sympathetically about COBOL is an act
bordering on heresy.”
Historical Perspective
Technical Perspective
Social/Psychological Perspective

56. Historical Perspective

57. Historical Perspective
“academic computer scientists did not participate in the design team”

58. Historical Perspective
“academic computer scientists did not participate in the design team”
COBOL contributors came from:
US Air Force
Honeywell
RCA
Burroughs Corporation
David Taylor Model Basin
Sylvania Electric Products
IBM
Speary Rand
Bureau of Ships
US Steel Corporation
DuPont Company
Air Materiel Command
NCR
USA - Signal Corps

59. Historical Perspective
“the COBOL developers apparently had little interest in the academic or scientific
aspects of their work”

60. Historical Perspective
“the COBOL developers apparently had little interest in the academic or scientific
aspects of their work”
CACM of May 1962 was dedicated to COBOL, only four papers had any
references.

61. Historical Perspective
“the COBOL developers apparently had little interest in the academic or scientific
aspects of their work”
CACM of May 1962 was dedicated to COBOL, only four papers had any
references.
But …
“Thus, of 15 papers presented at an Office of Naval Research (ONR) symposium
on automatic programming for digital computers in May 1954 [2], only two have
separate acknowledgements and none refers to other papers.” -- Backus,
“Programming in America in the 1950s- Some Personal Impressions”

62. Historical Perspective
“the decision of the developers to not use the Backus-Naur Form notation as the
metalanguage to describe COBOL”

63. Historical Perspective
“the decision of the developers to not use the Backus-Naur Form notation as the
metalanguage to describe COBOL”
COBOL Committee invented their own specification language

64. Historical Perspective
“the decision of the developers to not use the Backus-Naur Form notation as the
metalanguage to describe COBOL”
COBOL Committee invented their own specification language
BNF first used to describe ALGOL 60
“Specification Languages for Mechanical Languages And Their Processors …”
includes BNF, along with others → Gorn, ACM December, 1961

65. Historical Perspective
“A fourth concern is the process of describing COBOL to the academic and
industrial community”

66. Historical Perspective
“A fourth concern is the process of describing COBOL to the academic and
industrial community”
Translation: For a long time there were no good books “emphasizing the
conceptual foundations of COBOL”

67. Historical Perspective
“the people who might have accepted the title of computer scientist in 1960 were
not interested in the problem domain of COBOL programs”

68. Historical Perspective
“the people who might have accepted the title of computer scientist in 1960 were
not interested in the problem domain of COBOL programs”
COBOL was good at:
● data processing

69. Historical Perspective
“the people who might have accepted the title of computer scientist in 1960 were
not interested in the problem domain of COBOL programs”
COBOL was good at:
● data processing
Computer scientists were interested in:
● numerical analysis
● physics
● engineering
● systems programming

70. Technical Perspective
Successes

71. Technical Perspectives - Successes
Record Structure
● aggregation of dissimilar items
● hierarchy of names for fields
DATA DIVISION.
FILE SECTION.
FD SAMPLE-FILE.
01 SAMPLE-DATA.
88 END-SAMPLE-DATA VALUE HIGH-VALUES.
02 SAMPLE-NAME.
03 FIRST-NAME PIC X(8).
03 LAST-NAME PIC X(8).
02 DATE-OF-BIRTH.
03 YOB PIC 9(4).
03 MOB PIC 9(2).
03 DOB PIC 9(2).
WORKING-STORAGE SECTION.
01 GENERATION PIC X(10).
01 BOOMER PIC X(10) VALUE "boomer".
01 GEN-X PIC X(10) VALUE "gen X".
01 MILLENNIAL PIC X(10) VALUE "millennial".
01 UNKNOWN PIC X(10) VALUE "???".
* define record (data) structure
Input Record Format (SAMPLE-DATA):
|_ _ _ _ _ _ _ _|_ _ _ _ _ _ _|_ _ _ _|_ _|_ _|
| FIRST-NAME | LAST-NAME | YOB |MOB|DOB|
| SAMPLE-NAME | DATE-OF-BIRTH |
| SAMPLE-DATA |

72. Technical Perspectives - Successes
“[S]eparation of data definition from procedural
aspects”
IDENTIFICATION DIVISION.
* program name
ENVIRONMENT DIVISION.
* machine-dependent stuff
DATA DIVISION.
* define record (data) structure
PROCEDURE DIVISION.
* define operations

73. Technical Perspectives - Successes
Diverse set of control structures
● IF-THEN-ELSE
● PERFORM

74. Technical Perspectives - Successes
COPY keyword (kind of like “include”, copied content of a file into a program)
● organizational standards were easily enforced
● improved cooperation
● encouraged code reuse

75. Technical Perspectives - Successes
ENVIRONMENT Division
● specify machine or compiler dependencies

76. Technical Perspectives - Successes
Portability and standardization
● first truly cross-platform language

77. Technical Perspective
Failures

78. Technical Perspectives - Failures
No local variables
● required careful coordination of collaborators
● in 1974 CALL-USING was added, permitted parameters and runtime creation
of procedure names

79. Technical Perspectives - Failures
English-like languages
● supposed to be readable by managers
● “too wordy”, “somehow unscientific”

80. Technical Perspectives - Failures
Poor control flow readability
● originally, IF blocks were terminated with a
period (“.”), which was easily missed when
reading code
PROCEDURE DIVISION.
BEGIN.
OPEN INPUT SAMPLE-FILE.
READ SAMPLE-FILE
AT END SET END-SAMPLE-DATA TO TRUE.
PERFORM REPORT-GENERATION UNTIL END-SAMPLE-DATA.
CLOSE SAMPLE-FILE.
STOP RUN.
REPORT-GENERATION.
SET GENERATION TO UNKNOWN.
IF YOB >= 1980
SET GENERATION TO MILLENNIAL
ELSE
IF YOB >= 1946 AND YOB < 1965
SET GENERATION TO BOOMER
ELSE
IF YOB >= 1965 AND YOB < 1980
SET GENERATION TO GEN-X.
DISPLAY FIRST-NAME SPACE LAST-NAME SPACE GENERATION.
READ SAMPLE-FILE
AT END SET END-SAMPLE-DATA TO TRUE.

81. Technical Perspectives - Failures
Poor string handling
● EXAMINE, TALLYING, REPLACING
● in 1974 added INSPECT, STRING, UNSTRING

82. Social/Psychological Perspective

83. Social/Psychological Perspective
“fundamental differences between the computer science and business data
processing communities”

84. Social/Psychological Perspective
“fundamental differences between the computer science and business data
processing communities”
● dismissed the complexity of the problem of data processing

85. Social/Psychological Perspective
“fundamental differences between the computer science and business data
processing communities”
● dismissed the complexity of the problem of data processing
● had difficulty describing data processing problems with their theories

86. Social/Psychological Perspective
“fundamental differences between the computer science and business data
processing communities”
● dismissed the complexity of the problem of data processing
● had difficulty describing data processing problems with their theories
● “university professors did not like dealing with current practice”