.NET guys think using SQL tied to objects and stored procedures are good. This makes the application faster but less portable. The .NET guys say "hu, why should I write protable code?". J2EE guys think thats bad and prefer portable and maintainable code over raw performance.

I assume, because both technologies use the same basics (JIT etc.) they will have roughtly the same performance, especially with heavy HTML caching enabled. To run a performance benchmark, just rewrite the .NET PetStore in Java, which is pointless to the Java guys because they say: "I would never write such unmaintainable code." Can Java be faster (on reads) ? Perhaps with CMP strategies (read ahead, multi-fetch, deep-fetch). Perhaps not. Can .NET scale as J2EE ? Probably. Is it as easy to cluster a .NET application as in J2EE ?

What's funny is how TMC shot themselves in the foot. I laughed out loud when I read they have a LOC and have classes and methods they don't use. And .NET programmers call this benchmark realistic. Thats funny and with a laugh I might be more productive, this is the only thing that counts.