Where is Oslo going?
It’s always great to go through debates on new technologies. It is sure to give you some interesting insights into the technology at the same time forcing you to think n number of times whether you would want to invest your time on it. I came across such a debate on Microsoft “Oslo” today. There is a post - Where is Oslo going? at ebPML.org. If you are a developer who is getting started on Oslo, you might get disheartened when you read this one and you might start asking yourself whether you would want to really invest your time on this piece of technology. This post criticizes the thinking behind Oslo and its usefulness in moving towards Model Driven Development. The author of the article concludes by saying, “This project is focused on solving problems that people have already solved and completely missing the mark on MOP.” For those of you who don’t know what MOP is; it stands for Metamodel Oriented Programming.
But Doug Purdy (Product Unit Manager, Microsoft) decided to answer many of the questions that the author had raised in his post about “Oslo” and its relevance. He has put up a post on his blog to answer these questions. When you read this one, I am sure you will get some kind of confidence back on the technology. Going through the post will also give you an understanding about the problem “Oslo” tries to address which something like the DSL toolkit today doesn’t.
Interesting read!
Comments
Krishna:
I just want to make sure that I don't discourage anyone to try and use Oslo. Not everybody needs to create a Metamodel Oriented Programming Model. If all you care about is creating simple textual DSLs on the .Net platform this is the only game in town.
However, I suggest that some of the more ambitious goals around SOA may not be met due to a lack of architecture.
Posted by: Jean-Jacques Dubray | April 18, 2009 02:42 AM
Jean,
Thanks a lot for your comments and insights. I perfectly took your post in the right sense and understood what you were trying to convey. As you rightly said this is a first of attempt in the .NET world and hence looks exciting. As a developer I am more excited about the fact that the platform through MGrammar provides me the ability to write textual DSLs (need not be that simple) which was not something which was possible till now in .NET world. I see people using MGrammar (at least) in a lot of interesting ways to solve their domain specific problems. Now if we can store all the models (instances) into a single repository and can view/edit this model data through various channels then I think it would be useful in future. But I am not really sure about this second part as I still need to do some exploration on that front. With the launch of the M specification community, I feel that the M language will become more and more powerful and feature-rich in future. I think that we are seeing just the tip of the iceberg!
Posted by: E.Krishna Kumar | April 20, 2009 06:36 AM
Krishna:
Yes, it is exciting to have this capability in .Net as well. The Java world had this available for a couple of years already (openArchitectureWare).
Kris Horrocks wrote a follow up post to explain his basic assumptions when it comes to a modeling architecture. I have written a response here: http://www.ebpml.org/blog/188.htm explaining that "anemic" DSLs may not be enough. If you take HTML (Kris's example), HTML as a metamodel to construct web pages models (interpreted in the browser), well HTML is nothing without JavaScript. I think one should always keep that in mind when designing a DSL.
Posted by: Jean-Jacques Dubray | April 22, 2009 09:53 AM
I am still waiting to see a complete example solving a Business Problem (may be a very simple one).
I think it will be more clear once Quadrant, .NET4.0 and Dublin are released.
Till that time everything is looking exciting specially the repository which can hold even application definition along with the data. I am sure Dublin will use this repository too in combination with tracing functionality of its own to provide some really meaningful insight of application behavior.
Posted by: Ajay Singh | April 24, 2009 05:25 AM
Jean,
I went through your post and found it very interesting. I want to be clear in my mind regarding what you are saying. Are you saying that in the current state we cannot define behaviour through MGrammar? Are you saying that we need to have the capability to define behavior in the models so that we don't need "code behind" to drive our models because at present it looks like a way to convert unstructured data into structured data and then use that structured data in whatever way you want?
Posted by: E.Krishna Kumar | April 24, 2009 06:03 AM
Ajay,
You are right! I feel a lot of things are not very clear right now. Developers are using the MGrammar in different ways to solve different problems. But I guess that is the whole idea of having domain specific languages. We all agree that this is a powerful capability, but we need to wait and understand how best we could utilize the "Oslo" platform for application development and maintenance.
Posted by: E.Krishna Kumar | April 24, 2009 06:20 AM