OOP 2007: Vendors step into DSM

OOP 2007 showed me that now more vendors see DSM as the way to go.

Gentleware, known for their UML tool Poseidon promoted DSL-based editors built on top of the Eclipse GMF platform. When I asked Marko Boger about what happens when the customer needs the inevitable change to the supported DSL language his answer was: "Well, then we make more turnover". He could also have mentioned that the customer loses all models made previously and that the customer has to wait until Gentleware makes the new editor available to them. Not exactly the most flexible solution if you ask me, which might have been a reason for him not mentioning it. Still it is nice to see that someone who earlier concluded that UML profiles are sufficient now agrees that in order to generate code, developers need to work with domain-specific languages.

It was encouraging to see so many people participating both the MetaCase vendor session and Juha-Pekka's conference session. Especially the fact that some intelligent discussions took place about defining domain-specific modeling languages among the audience showed me that the audience has matured in this area. Where 3 years ago we had to explain to nearly each and every one what DSM meant now people we more interested in the "how to do it". They got the idea and agree with the benefits it offers, now they want guidance in how to get started with it. I see it as a big step forward.

Some of the best hitting comments I got from attendees were: "MetaEdit+ is the reason why Rational invented UML as their tool could support only one modeling language", "MetaEdit+ is MDA done right, 10 years ago" and "These people are the only ones who know how to do model-driven development correctly"

Webcast: Modeling with a Domain-Specific Language

We published a 12-minute informative webcast on domain-specific modeling. In it, Juha-Pekka directly compares UML with a DSM language-based approach. This makes the difference between them very clear.

Have a look here.

DSM Focus on Code Generation Event

Code Generation 2007 promises to have a significant focus on domain-specific modeling for model-driven development. With Microsoft and MetaCase already signed up as sponsors, I believe we can be quite sure of this.

The event organizers have managed to get some influential speakers in the area of MDD: Axel Uhl from SAP, Markus Völter, Steven Kelly (MetaCase), Andrew Watson (OMG) and of course the people from the Microsoft DSL team.

If I understand correctly, the program promises to give attendees a valuable learning experience with a rich mixture of sessions from 75 minutes up to well over 3 hours.

DSM growing, UML slowing

The number of hits on Google for the string "domain-specific modeling" is clearly showing exponential growth. Today Google treated me with 88,100 results, making the graph I created 2 days ago (above) already outdated (darn...it has that same characteristic UML models also have).

I was interviewed by the German magazine "ComputerZeitung" for their focus on modeling and model-driven development, the article is available here.

Reading it myself, it surprised me to have the magazine report Bran Selic talk about a new wave of UML:
"Wir bauen substanzielle semantische Definitionen ein. So wird die Sprache ausdrucksstärker und zu einem besseren Kommunikationsmittel."

Translation: We're building substantial semantic definitions into UML. This way, the UML will become more expressive and a better vehicle for communication. In other words, it seems the OMG thinks it a good idea to make it even more complex and even less suitable for code generation: Adding precise semantics to a general-purpose modeling language, leads you to some sort of ehr... general-purpose-domain-specific modeling language (hmmm try to market that).

UML is poorly suited for generating anything else from it than simply a copy of what it represents. Normally this means documentation. Adding precise semantics to the UML means the committee of tool vendors (all with their own agendas) need to agree on those precise semantics, which is impossible, period. The precise semantics will most likely end up being very loosely interpretable precise semantics, that you then can probably extend with semantic profiles of something like that. The end: a 160,000 page UML specification with the "U" clearly meaning "universal". Great idea Bran, let's add some more stuff to the UML.

The OMG seems to struggle with its mammoth specification, a new wave is needed because competing, disruptive technologies (domain-specific modeling) is eating away at the base. The only thing to do: move upmarket, add more stuff to your product (read this thought leading book on it).

A mark on the wall of UML's demise, is Telelogic, which now offers its UML tool for zilch, all you need is register and download. Their sales team will then no doubt hunt you for costly add ons. The UML tool market is under heavy stress, with hundreds of free tools which are perfectly fine for UML since you cannot generate all that much useful stuff from the de-facto modeling language anyway. Some tools claim they do however, and therefore cost mucho, you'll need to fork-out a substantial budget. Then, according to a Compuware-sponsored laboratory test (wow, must be very reliable) you could achieve a productivity increase of 35%. According to the Compuware marketing professionals this is a real "Landmark".

Funny that you do not see so many more of these studies isn't it? After all, MDA and UML are already so widely applied if you listen to the MDA tool vendors and OMG honcho's. Really, where are all the success stories of UML? Care to share one?