The idea that applications should be consistent across OSes is not really pointless, it just happens that for some reason the webbrowser is now one of the major platforms for applications instead of java running in the browser. Also in the 1990s there was not a clear indication that x86 would dominate the whole market and in particular for non customer facing applications there really was no compelling reason why they had to look good or "native", so Java certainly looked like a good idea, as did Smalltalk.

The incentive for Sun probably was in part that they wanted their high end workstations to stay relevant and maybe eventually develop cpu's that accelerated java execution.

Yet on mobile, we go out of our ways to make things look and feel native, even if the web's being used.

The concept of run anywhere, from a code execution standpoint, is fine. That doesn't preclude using platform specific APIs to make the application better. Like using C# but targeting iOS's UI toolkits. Or using signals from a Mono application.

While MS may have certainly intended to fragment Java, the core concept of platform specific APIs isn't inherently a bad thing.

Also note that during this time, MS supported Windows on non-Intel architectures so I'm not sure about the x86 comment.

How were they supposed to make money off it? It was supposed to help them sell servers (by making web programming and enterprise programming easier).

How does having a cross-platform system make their servers or systems preferable?

I'm all for it and see Java as doing quite a bit of good (despite the terribleness of the language and brain-damage it's inflicted on developers, which is a great bit of bad). I just don't get how it helped Sun, and I suppose, history shows it did not.

By increasing the ease of writing stuff that needs servers, the idea was that more servers would sell. Some (not all) of those would be Sun servers. That's a net win for Sun.