1968The conference that named it
The phrase entered general use at a conference in Bavaria convened to discuss the state of programming large systems. Its report was published and remains readable.
The name of the conference was itself an argument. The organisers chose a term implying that the activity ought to rest on theoretical foundations in the way established engineering does, and they chose it knowing it was provocative and not yet earned.
What the complaint actually was
Projects delivered late, over budget and unreliable, at a scale that was new. The specific observation underneath is more interesting than the complaint: machines had become far more capable, so the systems people attempted became far more ambitious, and the ability to build them had not grown at the same rate.
Stated that way it is not a failure. It is a gap between two rates of improvement, which is a permanent condition rather than an emergency.
1972The lecture that put it well
A Turing award lecture four years later made the same point sharply: as long as there were no machines, programming was no problem; when there were a few weak ones it was a mild problem; and now there are gigantic computers, programming is a gigantic problem.
The difficulty is created by the capability. Every increase in what machines can do enlarges what people attempt, and the enlargement arrives immediately while the ability to manage it does not.
Why the crisis never ended
Fifty-seven years later, surveys of project outcomes report the same categories of failure in similar proportions, and the complaints are close to verbatim.
Either the crisis has been continuous for longer than most careers, which makes the word wrong, or the framing was wrong from the beginning. A crisis is a decisive moment with an outcome. This is a description of the activity.
1986The essay that said so
A widely cited essay divided the difficulty into what is inherent to specifying and structuring a complex thing, and what is accidental, belonging to the tools and languages.
Its claim was that most of the remaining difficulty is inherent, so no single development should be expected to produce an order of magnitude improvement within a decade. The prediction has been argued about ever since and has not obviously been refuted, which is itself informative.
What the framing cost
Calling a permanent condition a crisis creates a market for cures. Every generation since has produced methods, notations and tools presented as the resolution, adopted widely, and followed by the same surveys reporting the same proportions.
That is not a claim that nothing has improved. A great deal has. It is a claim that the improvements have been absorbed by increased ambition, exactly as the original diagnosis predicted.
What the conference got right
Two things, both durable. That the difficulty is one of scale and communication rather than of individual cleverness. And that a discipline needs shared vocabulary, published results and the possibility of accumulating knowledge across projects.
The second is the one still not achieved. Most organisations learn the same lessons privately and repeatedly, which the parts of this document about repeated numbers and misread papers illustrate from another direction.
The word engineering, and whether it was earned
The comparison the name invites is worth taking seriously rather than as rhetoric. Established engineering disciplines have shared handbooks of settled results, licensed practitioners with defined liability, and codes derived from investigated failures.
This field has none of those in any general form. Failures are rarely investigated publicly, findings are rarely transferable, and there is no handbook a practitioner is expected to have read. Whether that is a deficiency or a consequence of the material being different is genuinely argued, and the argument is as old as the name.
What the comparison misses
The other disciplines build things whose materials have stable properties, measured once and reused for a century. Software is assembled from other software that changes under it, and a component correct today may be incorrect after somebody else's update.
That is a genuine difference and it is used as an excuse more often than it is used as an analysis. Both uses appear in the same discussions, and separating them is most of the work of thinking clearly about it.
Why the word persists anyway
Because it is useful to whoever is asking for something. A crisis justifies a budget, a mandate and an exception to normal process, and a permanent condition justifies none of those.
Which is worth noticing whenever the word appears in current usage about any part of this field: it is frequently a claim about urgency made by somebody who benefits from urgency, rather than a description of a change in the world.
What we cannot verify
The conference report, the lecture and the essay are published and can be read in full. Project outcome surveys are compiled by consultancies and industry bodies using definitions of success that differ between reports and over time, and comparisons across decades are not sound; none is quoted here. Whether the field has improved is not settled by any evidence we would rely on.
In short
- The phrase and the term software engineering were both chosen at one 1968 conference.
- The complaint was a gap between what machines could do and what people could manage.
- Capability creates the difficulty, because ambition enlarges immediately.
- Fifty-seven years of the same complaint means the word crisis is the wrong one.
- A permanent condition named as an emergency creates a market for cures.
- The undelivered part is accumulating knowledge across projects rather than privately.