More OCaml

Steve wasn't planning a second OCaml post; this one exists because, in his words, "gosh" — a word he says he uses advisedly. What holds the short rant together is the delivery: every fresh marvel arrives pre-shrugged ("Sure, yes," "it's no big deal that," "you might not even care"), a stack of feigned boredom that keeps climbing until it can't. It is a crush diary, filed a month after the first post, and it reads like one. The claim that code which clears the compiler "usually works correctly the first time" is the sentence every new ML convert writes around week four, which doesn't make it wrong. The wobblier move is the economics: the small-community argument only works if you grant the 10x productivity claim up front, and it would acquit any niche language ever built.

The page also carries its own rebuttal. Years later Steve prepended a note conceding OCaml has "fairly fundamental problems" for server-side work, demoting it to a C++ substitute for client-side GUI programs while maintaining it's still a really cool language. Reading the 2004 enthusiasm with that ceiling already visible is half the fun. The rest is the ending, where the feature list gets dropped entirely for a quieter question: what actually makes a programmer commit to a language once the benchmarks stop mattering.