SingletonInUmlForJava tells us something interesting about how people expect to use the UML.
----
''Why would one try to express Java-specific implementations of Singleton in a '''modelling''' language? ''

	*To understand the limitations of the expressivity of the language.
	
If I'm going to use UML, and take it seriosuly, I'm going to spend some time nibbling away at the boundaries, understanding exactly what it is that UML can say and what it is that UML says well. 

An example from AI: A large part of my recent work has involved formal descriptions of inference algorithms as components (for reuse). We (my collaborators and I) have been working on a much richer notation than UML. What we'd really like to do (and I have done this a little bit) is use UML to explain our modelling language. This is a lot like what you're suggesting (in terms of 
stereotypes et al) but requires that we fully understand where mismatches occur (e.g. we need to understand the fringe case very well). 

WilliamGrosso

	*These things crop up in models.
	
If you buy the full Rational process, at some point, the model will start to take on some aspects of source code. And to the extent that it does, you need to be able to express language specific idioms.

	*The standard patterns should be easily manifest in a model. 

If the standard way to express a certain pattern is to annotate the model with text or stereotypes (which really turn out to be a fancy way of annotating the model with text), then does it' really facilitating communication among developers as much as is claimed? 

So how do people feel about the PartitionOfModelsIntoPicturesAndText


----
contributors WilliamGrosso, KeithBraithwaite