Lisp Wins (I think)

Steve opens with a confession: after seven years and half a million lines of Java, he's spent a year quietly hunting for a replacement. The core of the essay is a small comparative study — print the squares of 1 through 5, separated by spaces. Perl does it in two lines, Ruby in one, Python in one, Java in roughly twenty; the Java 5 cleanup with autoboxing and the "smart" for-loop saves almost nothing. The verbosity is a symptom; the disease is that Java offers no way to make your own shortcuts. Without macros, without templates, without a way to lift recurring patterns into the language itself, the boilerplate accumulates forever. From there Steve walks the metaprogramming landscape — C++ templates, AspectJ, and operator overloading (each a partial answer, too verbose for the power), Ruby and Python's MetaObject Protocols (semantics only, not syntax), Smalltalk, OCaml's macro system — and lands on Common Lisp as the only language where syntax is genuinely up for negotiation, "the gold standard" where good programming is refactoring via macros.

The closing caveat — appended after the essay went up — is the most-quoted bit: Lisp may win on language, but Java still wins on platform, and Steve is, at the moment, uncommitted. The piece is Steve's most candid statement of the choice he was wrestling with in 2004, written in the gap between two careers' worth of Java code and the decade of writing about why something better was still possible.