Royce 1970 and the Waterfall Misreading

Narrative

Independent writers converge on one core reading: Royce's August 1970 WESCON paper never uses the word "waterfall", and Figure 2 is followed almost at once by his remark that the implementation is "risky and invites failure". Practitioner blogs (Pragmatic Tobi in 2012, Lou Franco, Sam Hogarth in 2023, Andreas Zwinkau, the bawiki essay) all rest on the same close reading. Several add the detail that Royce's Figure 3 allows iteration only between adjacent phases, and that he treated long backtracks, such as test failures forcing a redesign, as the real danger. Martin Fowler's bliki makes the same point, which is why "Royce was an iterative thinker" is partly supported and partly oversold.

The naming and origin story is where the better independent sources diverge from the popular retelling. Bell and Thayer's 1976 ICSE paper is the earliest known use of "waterfall" for Royce's activities, and a 2023 arXiv simulation paper confirms this by quoting them. Laurent Bossavit's "The Leprechauns of Software Engineering" (chapter 7) is the strongest sceptical source. He reconstructs citation history through Google Scholar and argues that both camps are wrong: the "agile revisionists" who say Royce advocated iterative and incremental development, and the defenders who say he was misrepresented. He also shows that early authors often wrote "Boehm's waterfall", and treats Boehm's 1987 remark that prototyping is "compatible with" waterfall as the moment the model split into a "stupid" half and an "enlightened" half. Bossavit's own point is that "Royce invented waterfall" was a convenient myth. The bawiki essay and the Hacker News thread dispute the strong form of the myth from the other side: single-pass development predated Royce, and Figure 2 is what people mean by the word.