Would it not be smart for MS to take up UE3.0 development and create platform-specific optimizations? Could the ongoing UE3.0 development fit into such a development strategy, with generic UE3.0 for the masses, and an MS sponsored tailored engine for their system?
What ongoing UE3 development? Last I recall, UE3 is pretty much in the "support and continued bug/usability updates" phase, and UE4 is the real growing project. And the short answer to your question is "yes, it would not be smart." Mainly because everything they'd do will most likely be removed by the licensees and replaced with something that suits their purposes.
The longer answer is quite simply the fact that there's no such thing as a one-size-fits-all engine, and the lower level optimizations are where there tends to be a lot more divergence, because the needs on that end tend to become project-specific, and there are combinatorically many possibilities. No one ever uses an engine as is unless they're in the business of expansion packs and mods. For instance, say you're a licensee don't WANT to use the shadowing method in UE3 and roll your own, what good does it do for MS to optimize the hell out of that when someone is going to rip it out? Say you have a shader-ized AA fake that looks better on average than hardware 2x or 4x MSAA (which is really not hard), then nothing good can come out of having an engine optimized for tiling -- if anything, it will hurt you. Say you have a hefty solution for vegetation collision, you're basically going to write stuff in on your own that somehow has to mesh with the existing API, and if someone has done more damage to the low-level parts of the API, you're more likely to run into surprises, and surprises are never in your favor.
This would be similar to PSSG being a Sony sponsored, cross-platform but PS3 optimized graphics engine, only with the reach of the dozens of titles using UE3.0. If you can offer an optimized engine that showcases your hardware over and above the unoptimized engine on your rival, especially when 1 in 3 titles use it, that has to be a smart use of development time.
I can't say I've looked even the slightest bit into PSSG, but FWICT, I wouldn't call it an engine in the same way that I'd call UE3 an engine. And if you read the descriptions, it's not a full game engine the way UE3 is -- it's strictly about rendering. Essentially it's a render layer and possibly some corresponding asset generation tools and libs for a game engine to use. That's part of the reason PSSG works as a concept -- it's only one susbsystem, and specifically, it's the last one in the chain of dependencies.
Now Microsoft doing something closer to that, I can see as being semi-useful, but because Xenos by nature has some more idiosynchracies to it, ultimately, I'd see it becoming more of a development guide than a development tool.
