"One of the difficulties in thinking about software is its huge variety. A function definition in a spreadsheet cell is software. A smartphone app is software. The flight management system for an Airbus A380 is software. A word processor is software. We shouldn't expect a single discipline of software engineering to cover all of these, any more than we expect a single discipline of manufacturing to cover everything from the Airbus A380 to the production of chocolate bars, or a single discipline of social organization to cover everything from the United Nations to a kindergarten. Improvement in software engineering must come bottom-up, from intense specialized attention to particular products."

“We follow two rules in the matter of optimization:”

“After forty years of currency the phrase "software engineering" still denotes no more then a vague and largely unfulfilled aspiration.”

“A development method may be regarded as a path or a procedure by which the developer proceeds from a problem of a certain class to a solution of a certain class. In trivial cases, the method may be fully algorithmic; for example, there is an algorithmic procedure for obtaining the square root of a nonnegative number to any desired degree of accuracy.”

“My first serious programming work was done in the very early 1960s, in Assembler languages on IBM and Honeywell machines. Although I was a careful designer — drawing meticulous flowcharts before coding — and a conscientious tester, I realised that program design was hard and the results likely to be erroneous. Into the Honeywell programs, which formed a little system for an extremely complex payroll, I wrote some assertions, with run-time tests that halted program execution during production runs. Time constraints didn't allow restarting a run from the beginning of the tape. So for the first few weeks I had the frightening task on several payroll runs of repairing an erroneous program at the operator’s keyboard ¾ correcting an error in the suspended program text, adjusting the local state of the program, and sometimes modifying the current and previous tape records before resuming execution. On the Honeywell 400, all this could be done directly from the console typewriter. After several weeks without halts, there seemed to be no more errors. Before leaving the organisation, I replaced the run-time halts by brief diagnostic messages: not because I was sure all the errors had been found, but simply because there would be no-one to handle a halt if one occurred. An uncorrected error might be repaired by clerical adjustments; a halt in a production run would certainly be disastrous.”

“The beginning of wisdom for a programmer is to recognize the difference between getting his program to work and getting it right. A program which does not work is undoubtedly wrong; but a program which does work is not necessarily right. It may still be wrong because it is hard to understand; or because it is hard to maintain as the problem requirements change; or because its structure is different from the structure of the problem; or because we cannot be sure that it does indeed work.”