Lisp Wins (I think)

The verdict at the end of Steve's year-long hunt for a Java replacement, first posted in August 2004 with the hedge built right into the title. Its most endearing habit is correcting itself in public: a bracketed note added a year later concedes that Ruby's metaprogramming turned out to be stronger than the essay gives it credit for, and Steve admits mid-argument that his tidy Perl example was a lucky find, "pretty unusual for Perl." The case underneath all that hedging is sturdy and funny in the details. Reshaping C++ with templates gets compared to "pushing a dead horse down the street," and there's a bit of Fred Brooks arithmetic holding that compressing 450,000 lines of Java into 100,000 lines of something denser should shed about 80% of the bugs.

The page preserves exactly one reader comment, from September 2004, and in hindsight it lands hard: what do you actually do with Lisp, the reader asks, when it has no libraries for networking or much of anything else? The essay offers no answer, and that question is the one "Lisp is Not an Acceptable Lisp" (linked above) spends 2006 conceding at length. The objection that eventually unwound this essay's conclusion was sitting quietly at the bottom of the page almost from the day it went up.