Yacc Is Not Dead: Why a Provocative Paper Missed the Point

The fuss

It has been reported that an arXiv paper titled “Yacc is dead” by Matthew Might and David Darais set off a flurry of online debate about whether traditional parser generators should be retired. The internet obliged — as it does — with hot takes and tribal lines drawn. But one clear voice pushed back: a detailed rebuttal argues that the question people want answered isn’t whether yacc should survive, but whether the new paper actually brings solid theory to the table.

Theory matters

The rebuttal alleges the paper ends up as a different kind of cargo cult: dressing up shaky ideas with the language of theory without delivering the underlying rigor. That’s not just quibbling over style. The author points to historical lessons — grep and yacc were successful because they were grounded in real theory (Brzozowski’s derivatives, Thompson’s NFA trick, later DFA caching). When theory is misapplied, you get real harms, like the exponential-time disasters seen in naive regular-expression engines. Ouch.

Old ideas, new framing

Yes, you can define derivatives for context-free grammars and frame parsing as successive derivatives — that’s been observed before. The kicker, though, is the practical difference between automata for regular languages and parser states for CFGs: NFA-like states tend to be lightweight and de-duplicable; parser states carry context and don’t collapse so neatly. In short: the math might extend, but the performance and engineering trade-offs don’t vanish. Surprise? Maybe a little.

The takeaway

So is yacc dead? Not on this evidence. The story’s emotional core is frustration — with sloppy rebranding of old ideas and with conversations that skip the hard questions. Want better parsers? Great. Show the proofs, measure the costs, and don’t skip the plumbing. Theory isn’t a cape you put on to make code fly.

Sources: research.swtch.com, Lobsters