I read a nice article in Information Week today.
IBM And Microsoft Have Dueling Visions For Software Modeling
This short article gets to the heart of the Oslo matter, which is Microsoft’s full court press initiative to provide modeling as a mechanism of building software and managing operations. I predict the most controversial part of Microsoft’s Oslo play will be in the modeling language itself, which will no doubt have little if anything to do with UML.
So, here we go again, right? YA industry standard ignored. Here’s my take: SO WHAT!? In fact, why not?
Over the years, I have come to the conclusion that "industry standards" are often propped up by other vendors, such as IBM, Oracle, Sun, and a standard cast of others. How is this cabal behavior different than anti-trust behavior? Smells similar to me.
Further, Microsoft has a successful track record of usurping standards with richer functionality not afforded by standards. Anyone remember XML Islands in IE4? How about DHTML?
My Predictions About MSML (Microsoft Modeling Language (I made that up)):
- It will be cool.
- It will take until 2013 to become practical (v3).
- It will spawn a whole new argument on standards compliance that will be as loud as TDD fights.
- It will have awesome designers and rendering technologies for the models.
- The ALT.Netters will hate it.
- It will become ubiquitous, in ways UML never did.
- UML will become cute. Like Lisp.
- It will fail to raise the genuine focus of developers to the model and away from code. Like UML.
Flame away.
I think the strongest point for UML is it is not vendor specific. Modeling makes sense when you can abastract the model from the code and talk about the essentials of the software no matter what is developed in. You can see patterns and apply patterns to code and systems.
MSML will be vendor specific, and while that is helpful if you only live in a MS world it is not realistic in Enterprise software. That is why I think UML will survive. It is capable of modeling the real enterprise software environment that is not vendor specific.
The more I learn of Lisp, the more important and less “cute” it becomes. But I don’t want to freak anyone out here, so I’ll point to this article by Don Box (who is awesomely smart and works for Microsoft) from the MSDN Magazine:
http://msdn2.microsoft.com/en-us/magazine/cc163729.aspx
I hope that these sources are deemed credible enough to read the article with an open mind. 🙂
UML will be trashed for MDD/DSL. New modelling language will be a re-entrant artifact which stays in sync with code (it IS the code), where as UML is troubled because it is seperate and always seems to veer off on its own path without regard to what the code is really doing, esp in agile.
UML will be changed to WML- Whiteboard Modelling Language 🙂
@Mike
Although I like Don and agree he is smart, if you want to talk modeling and quote an authority I am sorry but for me you will have to pick one of the 3 amigos and preferably Grady Bouch. There is one smart dude.
@Jarod
I am not sure where the perception has come from that UML does not allow good round trip Engineering. Take a look at the Enterprise Architect product line from Sparx Systems. Full roundtrip engineering, integration into eclipse and Visual Studio and support for multiple languages and platforms not just .NET. All for under $300 for a Enterprise tool This is UML based and not vendor specific. It provides code generation along the lines of the DSL tools MS has. Very cool.
Daryl, we have EA at my work. I am ‘somewhat’ familiar with it and dont particularly care for it. ‘Import/Export’ template generation is not the same a MDD->DSL. I will say I have not had the time to seriously sit down and grok the tool, mostly because I am turned off by it.
The UML PIM Camp vs Model Driven Development
http://martinfowler.com/articles/mdaLanguageWorkbench.html
I am not advocating a ‘skilled business line manager’ changing the architecture with a model on the fly, as the original article that Dave posted suggests MS is aiming for. But at the same time, Im sure you can agree, defining a successful business process ‘model’ is far more important to the success of an enterprise LOB app than the technology that does the work.
@Jarod
I agree some of the templates are the weakness of EA (although I undesrstand it is better in 7.1). Too say they are not the same as MDD->DSL is accurate. By definition MDD is much closer to the specific language you are using. However, all that aside once the import is complete (or your model definition is at the point where you can generate code) the roundtrip enginerring between EA and the language of choice is pretty much automatic.
My problem with this is that kind of defeats the purpose of the model which is to abstract the solution at a higher level to take a look at the overall patterns and architecture.
As for Folwers comments I agreed with him when his article was written 3 years ago. Now I belive the languages and tools have invalidated his arguments against UML PIM.
In order to define a successfull Enterprise LOB you must take into account all architectures that could be helpful to you. At that level a Language Specific modelling tool like a DSL is too close to a given architecture and solution to help. I think the abstraction of UML helps define better applications.
BPML and BPEL are the tools that help define a better busines process. I think EA has done a good job on incorporating those.