The world's biggest software company
is gearing up to tell you how its new modelling platform will transform developer productivity. Modelling is meant to make programming easier, by expressing the design of an application in easily understood diagrams rather than in a bunch of complicated C++
The snag is that while models are a great way to design, describe and communicate how an application works, in the end someone still has to write the code. Smart people have made immense efforts over the years to fix this by integrating the code with the model: generating code from models, generating models from code, or even devising ways to execute the model itself.
While these efforts have not failed completely, model-driven development has never taken off in a big way. Developers figure that they have enough trouble writing and debugging code, without taking on the burden of keeping a complex model in synch, or learning the mysteries of Model-Driven Architecture (MDA), with its PIM, PSM, MOF, OCL, ODM, SPEM and so on - have a look at the Object Management Group
(OMG) if you want to know more.
Now Microsoft is having another go. Code-named "Oslo", the project includes a store, a language and various tools; and it's going to figure in a big way at the company's Professional Developer Conference
next month. Some details are already trickling out. Posts from Don Box
and Doug Purdy
explain it in simple terms, and this eWeek article
has a bit more. The modelling language is code-named D, and I suspect it is another role for XAML - see my post on 10 things you might not have known about XAML
if you thought it was just for building GUIs and Silverlight applications.
It would make sense for Microsoft to use XAML, since it is a flexible textual language that is designed to play nicely with visual tools. I'm guessing that you will be able to create the model visually and then run it, just as MDA promises. In a video
, Senior Vice President Bob Muglia says:
When a model is created, it's possible for the business process to run directly with very minimal procedures being created around that. The model is the centre of the universe.
There are also hints that Oslo will tie in with Microsoft's cloud computing initiatives - maybe Live Mesh
or some more business-focused variant.
Will it work? Microsoft is certainly serious about it. It has even joined the OMG
, despite years of antipathy, though I hope this does not mean Oslo will be as complex and unsatisfactory as MDA. I see it as model believers getting together, and a way for Microsoft to get more academic and enterprise credibility for its initiative.
There are however plenty of reasons to be sceptical. Microsoft has had numerous modelling initiatives over the years. In Visual Studio 2003 it was Visio and UML. In Visual Studio 2005 it was class diagrams, Domain Specific Languages (DSL) and Design for Deployment, a way of taking deployment constraints into account before and during development. This article
explains the strategy, and enthuses about software factories as the way forward. Now there is Oslo, which looks like a brand new approach. Why think this will succeed any better than what has gone before? Why think that Microsoft will achieve what the fine minds at the OMG and its industry partners have failed to do?
There is also little evidence that this is driven by customer demand. When I talk to Microsoft developers, I don't hear many asking for better modelling tools; and if the subject does come up, it is usually a request for more standards-compliant and up-to-date UML diagramming, not full-on model-driven development. This is a top-down initiative.
Count me a sceptic then; but I also want to give Oslo a fair assessment when it is unveiled next month. It is quite simple. If Microsoft can deliver a modelling platform that really does simplify rather than complicate software development, and one that does not impose a heavy performance burden, then developers will love it. The big question: what is the chance of that?