Comments

• bruce January 14, 2005 5:47 AM

  Bruce, you said:
  > After all, the experts aren't constrained by anything when it comes to describing their world outside the realm of software. They have white board, diagramming tools, everying that a designer would need. If that was the way they wanted to go, they'd already be there. And yet they're not.
  
  I'm not sure this is generally true. I have seen business SMIs (subject matter experts) very happily draw diagrams, such as process diagrams. They also happily draw boxes to represent resources such as people (i.e. actors), systems, organizations, and entities (such as "customer" by which they mean their own record of a a customer). About 40 years ago (good grief! is it that long?) I designed business systems with SMIs who were very happy with flowcharts to specify processes. (This was before I discovered computers and fell in love with software, when the practice of design of non-computerized "systems" went under the name of "organization and methods".)
  
  Coming bang up-to-date, Richard Pawson has had significant success with an approach he calls "naked objects" (see www.nakedobjects.org), where SMIs define entities and lower-level processes or procedures, they are coded in Java, and the naked objects framework automatically produces a working visualization of the objects (a process such as filling in an order form being seen (rightly imho) as a process specification). The SMIs apparently lve this. Again, working with Haim Kilov (who wrote "Business Specifications" and other things) I saw SMEs, after about 20 minutes of education by Haim, happily using - on a whiteboard - a very small subset of UML that Haim had (carefully!) defined.
  
  So I conclude that SMEs are happy with diagrams, that sometimes they do use their own self-developed techniques, and where they do not already have a preference, wlcome the introduction of a *SIMPLE* techique to express their expertise.
  
  Bruce, you also said:
  > I think this is one case where we need to listen to the experts and find a way to represent the mental model that they have spent years developing.
  
  Hear hear! The key point it seems Lloyd Fischer was making is that we should not be presenting diagrams of software to SMEs. We should be helping them to specify their area of expertise. The trick then is to do so in such a way as to make the resulting design expressable in software. I have been involved in work along these lines for the past several years, and there is indeed light at the end of this tunnel. Several practitioners are using this already (e.g. COSM - see http://www.herzumsoftware.com). Part of the key is to use a component-based design approach for the application software. Not the "anything that sells is a component" approach, but the software engineering approach where you realize your designs in the kinds of components supported by COM, CORBA CCM, and EJB.
  
  
  

• bruce January 21, 2005 7:25 AM

  Actually I disagree that we need to help them specify their area of expertise. In many cases, they already have a method of specifying that information. The example I'm drawn to would be accountants. If you've ever discusses the transactions that need to be done in order to book depreciation across multiple states, odds are pretty good that the numbers person didn't turn to a UML diagram. In my experience, they fill the white board with T-accounts. I think what we need to do, and the reason why DSL should not depend upon UML, is find a way to allow them to use whatever notation they are comfortable with. Will some SMEs uses UML? Of course. But don't force those that don't think that way into the same box.

Leave a Comment