Politics between SUN and MS was the main reason. But the low quality (and by that i mean things like doing value type allocations in the heap, shitty libraries for UI (Swing), the language not having common sense features such as delegates/properties etc) implementation of the JVM was another.
There are many reason's it made sense for MS to create their own Java-like system, but the "low quality" of the JVM implementations is not one of them. The JVM implementations (HotSpot, IBM J9) are more efficient than the CLR in virtually any sensible benchmark you care to run and it's been that way since about December 1998.
e.g.
http://shootout.alioth.debian.org/u64q/benchmark.php?test=al...
That shootout is between the JVM and Mono. The CLR outperforms Mono.
That being said, my understanding is that JVM programs slightly outperform CLR programs in terms of speed while CLR programs do better than those on the JVM(s) in terms of memory.
the benchmark is pretty much irrelevant, because if it was we would all be coding in C or Fortran, or why not machine code or assembly ?
On the other hand I find that the mono project (young, open source implementation ), is doing quite great against java 7 (commercial implementation, unless I misread). It would be interesting to see how the openJDK is performing ? (the real question would be : is openJDK crippled compared to the Oracle - JDK ?)
>the benchmark is pretty much irrelevant, because if it was we would all be coding in C or Fortran
The point being debated is that the JVM is being developed and advanced more rapidly than the CLR. Performance is one way to demonstrate this. Performance does matter in many cases, and most folks aren't prepared to write their web services in Fortran.
>is doing quite great against java 7 (commercial implementation, unless I misread)
I would hardly classify 7x slower as "quite great"
>It would be interesting to see how the openJDK is performing ? (the real question would be : is openJDK crippled compared to the Oracle - JDK ?)
>>I would hardly classify 7x slower as "quite great"<< When half of that difference comes from running GC on a separate core, and we know that GC has not been tuned - perhaps we should call that particular difference accidental ;-)
