Okay, so this is like kicking a dog with three legs, but I cannot resist =) Apparently, Java is good enough. Michael has some very good points -- the biggest being that languages have become dominant in the past when they have ridden a new application paradigm. Java rode the move from client/server to... a different client/server with a web client backed by a gargantuan app server playing the same role as the mainframe used to, or as he points out, with the mainframe still right behind it.
Apparently Java is good enough -- specifically in the arena of web app development. I think he misses Java's real strength which is that there is a huge pool of mediocre developers available who can do good enough work when given tools indistinguishable from magic =) This is a huge benefit, which is not to be underestimated, and no, my tongue is not in my cheek when I say that, really. The productivity level achievable with IDEA (or Eclipse, or NetBeans) and Java for the Corporate Developer (Sun's marketing term for who they aim Java at, at least the one they say to people) is immense.
So, because we have tools which make the language irrelevant, is it really Java we care about? I think we need to stop calling ourselves Java Developers and start being honest about it, we are IDEA Developers, Eclipse Developers, or, to be completely honest, XMLSpy Developers.
You master your tool and can work wonders with it. I have macros,
er, I mean Live Templates, for probably a third or more of the code
I actually right. tnuo[tab], test[tab], set[tab], tear[tab],
xsde[tab], puts[tab], itar[tab], fore[tab], itli[tab],
etc[tab]. Nice. Now, consider for a second an alternate
approach. Consider a tool which made it really easy to make those
macros, by say, interpreting them at runtime. You'd have to pick
better combos than tnuo (throw new
UnsupportedOperationException("come back and implement
me!");) but funny how it works. Now, that could be kind of
handy. What if you can do the same thing for how the tool
works?
Sure, Java is good enough. So are two penny nails. Unless you are trying to eat ice cream. Then a spoon helps.
Ruby isn't going to replace Java. Java is a marketing term for Gnome. Ruby isn't a window manager. Well, er, it does play nicely them though, not too bad, actually. Sorry, back off the tangent. Ruby isn't going to replace Java like Java replaced C++. Java replaced VB (which is an awesome tool around a ... hey, wait a minute!).
Java is definately good enough. The key thing is for it to evolve to stay good enough. Will it? Well, the language is under the control of a committee of committees of people who dislike each other. Luckily, we aren't actually Java Developers, we are IDEA Developers, and IDEA is under the control of some very big-thinking Russians. I think us IDEA Developers have some potential, not sure about the Eclipse folks though, they have the whole open source thing going, but it seems like most of the control over Eclipse is by people who sell competition to it and use it as a testbed for new research ideas ;-) So it will either be an awesome amalgamation of cutting edge software engineering research and heart warming open source community goodness, or a bike shed. Overall, not bad. I think us IDEA and Eclipse, and even the NetBeans Programmers, have a future.
Seriously now, IDEA is a fantastic, er, language (?) when you have big complex systems with a significant number of developers. Static typing, good garbage collection, a good compiler, excellent code navigation, and medium-level performance for the actual runtime (note that performance != scalability). It has fantastic libraries via it's Java runtime system, and bindings to most every information system around. IDEA has legs.
I still prefer Ruby, personally, but I do more work in IDEA than I do in Ruby, so take that as you want.
Side not: Fortran still being used because it is "good enough" is bollocks. Fortran is still used because nothing is faster when you need hardcore number crunching. Part of this is the immense optimization that has gone into compilers, the very mature vector processing libraries, the harsh limitations in what you can actually describe in the language in order to help the compiler optimize the code better (us IDEA developers have complained about non-reentrant EJB's, how about non-reentrant *functions*) etc. Most scientists I have met (this is not a scientific survey, I don't know that many) prefer Perl. Go figure.