The article misses the point of backward compatibility for java. Consider a company that has desktop (Swing) applications that were written in the 1.2 or 1.3 era. That company can upgrade to 1.6, and use 1.6 applications alongside those 1.3 applications. However, a java that wasn't backwards compatible would either have to keep some way to identify legacy code and run the old JVM, or they would have to force the new language into the backwards compatible JVM, where cruft has also developed.

BTW, I haven't looked at python 3 yet. How have they adapted to two pythons on the system?