A classifier is a classifier.
An attribute/property/pin/part/pineapple/frangipani is an informational feature of the classifier.
It is "convenient", within a context to give such feature a "prominent" name.
There are 10 types of things about a classifier: attributes (informational features) and behaviors (things you can tell an instance to perform).
There are no other things.
No matter how you dress up reality with this, that or the other "approach" to modelling the universe.
A model is a picture of the universe that helps explain an aspect of something in the universe.
A house plan helps explain the dream between the architect and the builder.
A color patch helps (occasionally to) explain what the wall will look like when it is painted.
A UML diagram should be constructed to explain, within some conversational context, some aspect of a system.
If the so called "rules" of UML don't let me do that, then I break them.
Let me say this only once (for the second time)! The model is not reality. The house plan is not the house, the color patch is not the paint. To paraphrase(?) Sylvia Plath "A rose is a rose", a picture of a rose is not a rose but it may give some audience an idea of the plant that I am talking about.
Go think about a painting that you like, of a rose if you like, or a fish or a tree or a car or a scream. Why is it good? Because it engenders something in your psyche - it just "gels".
And explains.
Now consider the person who painted it. Were they worried about whether this shade of blue was called "Prussian" or "Cobalt"?
Or were they just using that shade to convey, as clearly as possible, to the "audience" some aspect about that flower that they wanted to explain?
Model to explain, perhaps even unto yourself, aspects of the universe. Dressing up models with OMG names and trying to comply with the so called "rules" is a waste of both your and your intended audience's waste of time.
Is a property the same as a "Prussian Blue"? I think not.
Not clear enough? Well let's go sit in a coffee shop tomorrow and you can explain to me "how to get from here to Tiffany's in time for breakfast", on the back of a napkin. Would I expect a Google map of the world? (Complying to dog knows whatever "standard" they use.) Now, given the current context, that we are sitting in a cafe in rural South Australia - would you try and draw a google map?
Classifier: Location
Attribute: geopos? (whatever)
Behaviour: Describe_route(destination::Location) as Scribble_on_napkin
bah! Sometimes I think UML modelling has got to the point of acute cranial-recto-insertion.
b
p.s. This rave was in part incited by a recent post on the dreaded "more than one instance of the same entity on one diagram" dinosaur. In particular let me refer to figure 15.1 in the most recent adopted specification. Oh dear! There is more than one instance of an Adapter Behavior in that diagram. Oh dear, worlds will collide! Oh no, don't tell me OMG just broke one of their own rules. Oh hang on - there is no such rule anyway. But why didn't they draw it with so many damn criss-cross lines as to make it totally unintelligible?
(Damn. It was another diagram, but the point is still there.)
