ISerializable in Israel

Continuing this thought, one of the more interesting answers that came up at teh TechEd Architecture panel by Guy Ron is that the Software Factories initiative is most likely to fail from the evidence thus far. He mentioned a big project for the healthcare industry (HL7 Software Factory) and how, even today, a long time after it started, it still has not finished. Microsoft was trying to build (with third parties) the full object model of the HL7 so that applications can be built against it.

For now, it has proved too complicated to come up with such generic components to be reused in “factory floor” applications, which may be a testament to upcoming difficulties that the software industry will face when trying to get on board with this methodology.

Guy mentioned that it may prove useful for only smaller scale projects and domain (like UI) and not for “real world” domains.

Q: Software Factories and UML

Has Microsoft Abandoned UML?

Is the Domain Specific Language path so different than UML?

Can they be combined?

From what I’ve seen so far, and from talks I had with various people (including Steve Cook, from the original UML Team, now at Microsoft), Microsoft have sort of “adapted” an easy to use and understand version of the UML. They took the basic idea of shapes and arrows and made it work on the level of a specific domain which is in turn used in a Software Factory.

Most people I know have done *some* UML, but most only know how to use maybe 3 or 4 types of UML Diagrams, let alone using UML Protoypes and such stuff. The main problem people face with UML is that using it to draw shapes, it then makes you give *specific meaning* to those shapes in your own domain using prototypes.

Instead of using proptypes on the UML (which are very time consuming) Microsoft has taken some of the simplest ideas and made them part of the tooling strategy. So when you draw a shape, you’re actually drawing some sort of *real* entity in your domain. You can then manipulate it and use it to build real code, and it (should) reflects changes in the real code.

The notion of a software factory is, to me, taking the UML shapes and giving them real meaning inside the tools. Not only that, to be able to create meaningful shapes for specific projects easily.

I still use UML to a degree – whenever I draw shapes on the whiteboard I draw UML. Whenever I want to make some interaction clear I draw some sort of interaction diagram. But that’s basically it. IN projects where you are forced to write so much documentation and use cases using UML it usually turns out to be a very complicated set of design docs that developers have trouble reading. When they finally get to reading the whole thing its usually way out of date anyway and one has to talk to the man who actually wrote the thing. See? *Communication* always wins.

Instead, what I try to do is have less documentation (butt enough to get things going) which can help start good conversations with the designer/architect in that project. In an XP project that’s could be the team lead.

So, in short – UML is too complicated, most people don’t know even half of it. Microsoft tries to make it simpler and more usable in the real world of its tools.

On the first day of Tech-Ed, on the assembly Talk (the one that takes two and a half hours), Steven Sinofsky, the new manager of the windows division, went up on the podium to talk. He was on after a couple of short films (whihc were funny) and a short introduction from Lior Tzoref (the official teched blogger who won't take comments). There was magic in the air as 3 thousand people watch Steve go up onto the podium to talk about his vision for a better windows.

This was the main assembly at the conference. Surely there were going to be some amazing demos of Windows Vista, XBox, or anything else that has the word "cool" as an alias these days.

Then Steve said something like "I'm pretty much new to Windows division, but I'm coming from Office, which I know much about". Silence.  I got the weird feeling that this talk was not going to be as great as I'd hoped.

Then Steve said something like "so let's talk about the new XML features in Office Server intropdkj standradslsdfjnsdfsc...". That was not a typo. There was so much noise in the great big tent where we were sitting that I couldn't hear what he was saying. People were moving their chairs backwards and started leaving the tent.

It was awful. From a full tent of 3000 folks, probably less than 1,000 stayed on to hear Steve Talk. I was not one of them - I left with a good excuse. I had a talk coming up later that day that I wanted to get ready for, and this was simply a big waste of time.  

So why did people leave?

• The Israeli Crowd is a big mess of impolite rude people who don't have an ounce of patience to hear something they are not interested in. That said - their actions have reflected the awful truth - this was a very bad judgment call on Steve's part.
• You *don't*, you just don't, talk about Office and XML features in a main Assembly of TechEd, where "cool" is supposed to be uttered every 3-4 minutes by the attendees.
• You don't let people out during the main assembly.
• You don't save the Vista demos for the end of a 2.5 hour talk.
• You get more videos and "cool" stuff going.
• You make sure there is good wireless reception in the tent, so that people will at least check their emails instead of leaving searching for a hotspot.

• You expect a crowd of people, many of who are not that technical after all, to sit and listen to a talk about XML and office while there is a pool, hotspots and beautiful girls outside the tent?

For once, I agreed with the crowd and went on my way. Even so, that's 30 minutes out of my life I'll never get back.

• TestDriven.NET - The Addin I used to demo the code with
• Debugger Visualizer for DynamicMethod and MethodInfo
• Mdbg - IL Debugger with full source in .NET (more info)

• cl1p.net - Make up your own URL, paste in data, go to another machine, visit the URL, copy the data back. An easy way to use HTTP to transport data between computers.
• csUnit 2.1.1 BETA - This unit-testing tool for .NET now supports .NET 2.0 and Visual Studio 2005.
• Automise - VSoft has released their general-purpose automation utility, with a GUI based on the fantastic FinalBuilder. You can download a 30-day eval copy here, or license it starting at $195

• NUnit 2.2.5 Released
  This is an update to the recent 2.2.4 release, the first new production release of NUnit in over a year. See this blog entry for a summary of features and links to the release
• Ohad keeps making his Beat-Box Demo cooler and cooler.It's for Tech-Ed Israel and demonstrates creating a beat mixing application that uses EntLib for all it's backend stuff.  I saw this live and this is indeed a demo for the books. In fact, he should distribute it after Tech-Ed Israel with source, because this is one cool demo!