I was talking with java.net former editor Chris Adamson this morning (via email), and our conversation got me thinking about the future of Java. Last week's poll, which asked "What's your view of the emerging JDK 7?", elicited an enormous volume of commentary. A fair amount of the debate was related to Scala. Scala is indeed receiving increasing attention, as the current Alexa graph for scala-lang.org shows:

Prior to this year's JavaOne, at a java.net Community Leader's breakfast I attended, someone said that java.net isn't really "about" Java -- rather, it's about the JVM, it's about anything that can run in the virtual machine. That statement stuck with me, though until now, a couple months later, and after chatting with Chris, I hadn't thought of anything to say about it.

In the early 2000s I was working for a small company as architect / lead engineer / project team leader for a Windows based high-volume document archive and retrieval product. Back then, .NET was just coming into being. It strikes me now that the underlying .NET concept is actually quite similar to what the JVM has become: you have a software platform that runs a certain kind of byte code, and you can develop code that will run on the platform using a variety of higher level languages, with the developed programs reduced by a compiler down to the appropriate byte code. Come to think of it, haven't we always had that? Doesn't everything get compiled down to some form of Assembly language, ultimately?

The Java concept, of course, is quite different from .NET, in that Java is a "write once, run anywhere" paradigm, while a big (secret) objective of .NET (IMO) was to facilitate the transitioning of hoards of Windows developers over to the "next big thing" in the Windows world. An easy path into .NET reduced the number of lead engineers, architects, and Development V.P.'s who might otherwise have said "Hey, if our team has to learn something entirely new anyway, why don't we just switch to Java and have our product work on all platforms?" C#, after all, is Java-ized C++, isn't it? I vividly remember thinking "wow, Microsoft is really scared of Java!" when I read the first preliminary 65 or so page C# specification.

Anyway, I titled this little essay "Will the JVM Outlive Java?" because it seems to me that languages, overall, are less durable than the platforms on which they run. Among developers, language preferences, it seems to me, are a bit like music preferences. Most people tend to consider whatever they entered adulthood with as clearly the best. Interestingly, at the time when they are learning their ultimately preferred language, that language/technology is always a hot, new, cool technology. Just as the music they ultimately consider best was new and different and cool when they were listening to it as high school and college students. Both the languages and the music may incite statements of dismissal or derision from their parents' generation, so a bit of adolescent rebellion is also involved.

Note: I don't count myself among either of these groups. I like plenty of new music, and find new languages and language concepts quite interesting. But, I do have to admit that there really isn't anything that has quite the same feel for me as David Bowie's Hunky Dory and Ziggy Stardust. The imprint of what you really liked in your late teens and early twenties is quite pervasive -- kind of like just-hatched chicks imprinting on the first large animate object they see... Hmm, are we really that simple ourselves?

Anyway, there's another reason the platform tends to outlive the language. In the discussion in last week's poll, there was a debate between a development team leader who was managing hundreds of thousands of lines of Java code and another team leader who was working on projects basically from scratch, using Scala. So, scotty69 (the Java project lead) said:

Please help me: I have - say - 200.000+ lines of 3 years old Java code, mature, thoroughly tested, battle proof. My customer wants me to add feature X with about 2.000 lines of new code and and 3.000 lines of existing code to be changed, tight budget as usual. Do your really think that it's a good idea to to introduce Scala in this scenario to "extend Java classes, call methods and so on"? Do you have experience in mixing a large legacy Java codebase with new Scala elements? Can you tell me how the tool-chain could look like?

For an established code base, converting to a new language is enormously expensive -- basically, it's wasteful, stupid. That's why in the scientific/engineering data center where I work part time, we still have loads of operational Fortran libraries. Yes, we wrap them inside layers of other code, connect to a modern Oracle database, generate interesting graphics and post them on nicely formatted web pages, etc.; but the core computational code is legacy engineering and mathematical modeling code, mostly written in Fortran, or C.

Meanwhile, here's aehrenr's point (the leader of the Scala-based project):

We make applications for genome comparison including 3 tier apps with Swing frontends. And we do it as an university working group with changing programmers, many of them not computer scientists but also mathematicians, biologists and so on... I am for sure the exact opposite of a "java basher". Scala is starting to be used in the enterprise in reality...

If you're developing something from scratch, with relatively inexperienced developers, why not go with a well-constructed new language that solves some of the issues that are outstanding in older languages? If it's 1998 and you're developing a brand new application, why deal with stamping out C/C++ memory leaks and trying to teach young developers about pointers, when Java offers a garbage collector for free? Well, now it's today, 2009. Is there perhaps a new language that runs on the JVM that fixes some annoying, developer time consuming Java issues?

But, for this to be possible -- one set of developers working with established code bases that utilize rock-solid, well-tested code that is potentially decades old, and another set of developers working on brand new applications using new languages -- the underlying platform has to be durable and extensive and extensible.

Hence, my answer to "Will the JVM Outlive Java?" Well, Java won't die any time soon, but the JVM is the critical enduring platform, really. What do you think?