Pete Coffee has the right idea in this article - you have to draw the line properly when you decide what is and isn't "good enough". He thinks the industry move to Java (and it's bastard stepchild, .NET) represents "good enough"; I don't think so:

The languages that I have in mind are Lisp, Smalltalk and Java, which look more different than they are. With its garbage collection, its basis in symbolic expressions, and its ease of writing and reading new code at run time, Lisp is arguably the parent--or even the symbiont personality--of every modern scripting language. Take Smalltalk and give it a C-syntax makeover, send it to a spa for compilation to leaner and meaner bytecode, and you arguably have Java. Exploit fully the object model of Java, and you go back to the future: You find yourself using Java environments that have the same kind of introspective power and self-improvement capability that made me an ardent Lisper, and an aspiring (if never actually expert) Smalltalker, going on 20 years ago.

Leaner and meaner? Please. People continue to confuse faster benchmarks on arithmetic tests with faster in general. Java isn't "back to the future" - it's a roadblock slowing our progress. Can you patch a Java based application server in place without restarting it? I do that with this blog server all the time - I did it just this morning. Java isn't leaner and meaner - it's Smalltalk with a straightjacket, preventing you from doing anything that the designers thought was a bad idea.