Steve Cook vs. Michael Guttman on MDA, the short of it is Microsoft aren’t hot on MDA, but will do something that looks a lot like it. My initial reaction to this is: Vendors have little to gain from easing migration from their platform, so I can’t see them falling over themselves to help me produce a nice platform independent model. Call me cynical. However, vendors are going to take a great interest in improving my productivity as a developer on their platform, so models are definitely going to be important. Further thoughts on MDA:

UML’s concrete syntax is lousy – Not lousy in the sense that I can’t use it, just lousy once I get beyond UMLAsSketch. Model driven development is not going to look like UML – it is going to look like the problem I want to solve.

I don’t care if the meta-meta-model is lousy – I promise to not really care if my model is built on MOF or some other (proprietary) framework. Sure, I wouldn’t mind, but it secondary to what the model can do for me right now.

Is platform independence massively overrated? – Or, am I really sweating over PIMs and PSMs?

• I suspect the cost of moving platforms is much more than the cost of porting applications alone. Hence, in the enterprise scenario the “stick a web service wrapper on it and leave it alone ” approach will often suffice. On the other hand, I can see the huge advantages for anyone developing product lines and targeting multiple platforms.
• Developers will probably take the tool that makes them more productive right now, perhaps because that tool maps directly to their platform, over the tool that might save them time when they migrate to another platform at some later date.
• Splitting my platform specifics out of the model into a separate marking model isn’t a panacea if the marking model is a big part of the implementation. The relative sizes of the PIM and the marking model remain to be seen.

Updated:Comment from SDTimes.