They're choosing the Java Virtual Machine, not Java itself. Java just happens to the be standard example of a language which targets the JVM. Using the JVM is a prudent choice since many tools already exist which target this platform, it's sandboxed, and it's hardware platform-independent. You already have implementations of many languages targeting the JVM including Java, Python, and Ruby.
In fact, the JVM bytecodes are then recompiled to their special VM's object format, called Dalvit. Using the JVM as the common layer allows them to fit into the existing software ecosystem without causing pain for developers and language implementors.
Parrot always seemed like an odd idea to me, but I just assumed that there must be enough to be gained by doing such a high level VM if people were bothering to write it.
I was referring to performance and ease of adapting the JVM's way of doing things internally to very different languages. This is quoting a number of parrot and jython developers. I have the links around here somewhere..
Why do you think jython et all are so terribly far behind? It's not for shortage of interest, more because the likely possible outcome of this could only be sub-par.
I think Google picked Java because J2ME has already won a big midshare among mobile developers and Google wanted to appeal to them by providing a familiar tool.
Not sure exactly when that was, though, day and date.
(a) In the heyday of DOS, people used all sorts of languages. Some of the first DOS programs I wrote were in Prolog.
(b) If the best hackers don't want to build stuff on your platform, you're in big trouble. Hard as it seems to imagine now, DOS was impressive in the mid 80s, and smart people wrote apps for it.
I don't think Google is deliberately courting mediocre hackers by using Java. It was probably just the expedient choice for some reason.