THE CHRONIC SOFTWARE ENGINEERING CRISIS

1
THE CHRONIC SOFTWARE ENGINEERING CRISIS
By Walter Maner
2
DIJKSTRA
To put it quite bluntly as long as there were
no machines, programming was no problem at all
when we had a few weak computers, programming
became a mild problem, and now we have gigantic
computers, programming has become an equally
gigantic problem."
-- E. Dijkstra, 1972 Turing Award Lecture
3
THE CONCEPT OF SOFTWARE ENGINEERING

• Software engineering is
• the technological and managerial discipline
• concerned with systematic production and
  maintenance of
• quality software
• that meets the user's needs and
• is delivered on time and
• within budget.

4
BIRTH OF THE TERM "SOFTWARE ENGINEERING"

• The term "software engineering" was coined by a
  NATO study group in 1967
• Four decades later, it still hasn't earned a
  reputation as an engineering discipline

5
Software Engineering is LIKE ENGINEERING

• Both are dependent on advances in technology.
• Both are labor intensive.
• Both are pragmatic disciplines in the sense that
  they apply theory to particular situations.
• Both produce technological artifacts.
• Both rely on skilled teams.

6
Software Engineering is LIKE ENGINEERING

• Both require some form of analysis, design,
  construction, testing and maintenance.
• Both try to learn from failures.
• Both try to repeat procedures that work.
• Both use computer-aided tools (CASE, CAD)
• Both use engineering approaches to
  problem-solving (e.g., engineering notations).

7
Software Engineering is UNLIKE ENGINEERING

• Software regularly violates the law of
  proportionality that governs every other
  engineering discipline.
• Small changes may produce disproportionately
  large or even catastrophic effects.
• Advances is software technology occur at a pace
  too fast for us to master.
• Developers cannot rely on established best
  practices that have withstood the test of time.

8
Software Engineering is UNLIKE ENGINEERING

• During maintenance, software undergoes changes so
  radical that they alter the fundamental
  architecture of the system.
• It is hard to identify development practices that
  can be repeated with the same (or greater) level
  of success.
• Progress is hard to measure (but CMM tries).

9
Software Engineering is UNLIKE ENGINEERING

• Few good standard components are available
  off-the-shelf, forcing developers to custom-build
  many component parts.
• The technological artifact produced by software
  developers is nearly invisible during
  construction (prototyping helps).
• There are apparently no physical laws that govern
  the construction of software.

10
Software Engineering is UNLIKE ENGINEERING

• Lack of universal standards governing how
  software components should interface to other
  components.
• Time and costs are hard to estimate.
• There is no SINGLE technical notation that can be
  used throughout the entire process of software
  development.
• Software engineers are not bonded or licensed
  (but this is changing).

11
PreviewTHE PRODUCTIVITY CRISIS IN SOFTWARE
ENGINEERING

• Time To Develop A Program
• What Programmers Do With Their Time
• What Programmers Produce
• Mistakes Programmers Make

12
TIME TO DEVELOP LARGE PROGRAM

• gt 512K lines
• gt 100 programmers
• 48-60 months

13
TIME TO DEVELOP MEDIUM PROGRAM

• 64 K lines
• 5-10 programmers
• 12-23 months

14
WHAT PROGRAMMERS DO WITH THEIR TIME

• 13 writing programs
• 32 job communication

15
WHAT PROGRAMMERS PRODUCE

• Systems programmers
• 1 LOC/day
• Utility programmers
• 5-10 LOC/day
• Applications programmers
• 25-100 LOC/day
• Overall average for medium-sized programming
  project
• lt 10 LOC/day

Boehm, B. (1981). Software Engineering
Economics. Englewood Cliffs, NJ Prentice-Hall,
Inc
16
COST PER LOC

• 50 to 400according to Boehm and Standish, 1983

17
LOC GROWTHWindows NT
Doubling time 866 days Growth rate 33.9 per year
18
LOC GROWTHWeb Browser
Doubling time 216 days Growth rate 221 per year
19
MISTAKES PROGRAMMERS MAKE PER KLOC

• Errors found in code during development
• 50-60 per KLOC
• Errors found in delivered code
• lt 4 per KLOC (still a lot!)

Mills, H., Dyer, M, and Linger, H. (Sept 1987).
IEEE Software
20
SOFTWARE COST TRENDS RELATIVE TO GNP

• 1980 5 of GNP
• 1985 8 of GNP
• 1990 13 of GNP
• 1995 21 of GNP
• 1998 38 of GNP
• 2007 ?? of GDP

Cost total cost of ownership
Early data from Gibbs, W. Software's Chronic
Crisis. Scientific American. Sept 1994 later
data from Fortune Magazine
21
1979 GAO STUDY OF SOFTWARE BUILT FOR GOVT

• gt 60of government contracts had schedule
  overruns
• gt 50 of government contracts had cost overruns
• gt 29 of software was never delivered
• gt 19 of software had to be fixed in-house
  before use
• 2 of software was usable as delivered

Comptroller General. (1979). Contracting for
computer software development. General
Accounting Office Report. (FGMSD-80-4).
Washington, D.C.
22
1994 STANDISH GROUP SURVEY OF 352 COMPANIES

• 31 of all software projects are canceled
  before completion
• 53 of projects will cost 189 of estimates
• 9 of projects come in on time, on budget
  (large firms)
• 16 of projects come in on time, on budget
  (small firms)

http//www.standishgroup.com/sample_research/chaos
_1994_1.php
23
Standish Group Research Cost Overrun Data
24
Standish Group Research Time Overrun Data
25
Standish Group Research Dropped Feature Data
26
Standish Group Research Factors Making SD
Difficult
27
Standish Group Research Factors Making SD Fail
28
Standish Group Research Are We Improving?
29
Software Engineering InstituteCMM Data
The Capability Maturity Model for Software is a
model for judging the maturity of the software
processes of an organization. The model defines
five level of maturity and specifies what
processes that must by in place to achieve those
levels.
30
Data from Ed YourdonRise and Resurrection of the
American Programmer (1996)
The difference between the best and worst was
101 in 1990, but was 1001 in 1995
31
Standish Group Research New Data
32
Standish Group Research New Data
33
DEFECT COSTS

• The US Department of Commerce estimates that
  software defects cost the US economy 60 Billion
  per year

Infoworld, June 2002