Just got back from this years VBUG conference. It was pretty good two days, completely sold out and all that. Decent topic coverage too. For me personally there werent many new things cos i had the opportunity to go to the MSDN roadshow recently where we spent a whole day on WPF, WCF and WF with much longer and deeper sessions but hats off to the VBUG organisers for sorting out good speakers and topics (well, i cant speak for myself , i presented an Introduction to Biztalk, but i heard it went down well and the other chaps were a knowledgeable bunch.


There were a few glitches in the demos, but thats always the case at conferences. Even at mine, i ran out of time and abandoned the demo about 75% of the way through cos there were a lot of “lessons” that i need to share with the attendees and the demo was a simple file mapping one in any case. (Its the standard Biztalk demo, but it does show the abstraction at the heart of Biztalk so it suffices in most introductory talks). I stayed up late the previous night preparing some more comprehensive stuff including consuming web services etc, but couldnt show that. Still, i think we managed to cover a lot of ground in that one hour. I could probably start writing some articles based on the slides and some of my thoughts on using Biztalk with the other products in the MS Connected Systems strategy.


Its funny how perspectives change in a short time. There was a demo of PowerDesigner by a chap from Sybase. It was a good talk and he showed a lot of the features of the tool such as looking at the model round-tripping from UML to DB to code and back and things on requirements traceability, integration with VSTS etc. A few years ago i would have been on my feet cheering him on , but this time while i appreciated what he showed, i couldnt help thinking “its not TDD is it?”  and “how can we make this work with TDD? test, model, generate code, refactor, reverse engineer?nah.. too much work” which is rather ironic for me personally cos i used to be really hot on database modelling and code generation etc.  I’m still all for models and IMO you absolutely need the blueprint before you build the house, but the key point is how much of the blueprint do you really need before you can start delivering working code (and its different strokes for different folks).


I’m going to focus a lot on software factories and DSL’s and i believe theres tremendous payoff in that field and it can do wonders for productivity. More on this discussion of modelling and TDD later.