Ports Discussion

[davisford]

Regarding UML Ports, I am trying to figure out what the usefulness is.  Perhaps someone can provide some clarity.

Consider the following:



If that doesn't show up, try this hyperlink

http://photos1.blogger.com/blogger/6696/66/1600/ports.jpg
The first drawing is non-cluttered, and clear.

The second drawing adds a port.  If you add the name of the port, it gets more cluttered.  However, I fail to realize what additional semantics this really provides.  What is the point, really?

The third drawing aggregates multiple interfaces through a single port.  While this might be helpful to group interfaces, it gets cluttered and difficult to read.

Is there any other usefulness for the port.  I prefer the first style, but I am curious if anyone else has a different opinion.

[Paolo F Cantoni]

Davis,

Have you read the UML 2 Superstructure Specification on Ports?

That will explain why they are there (their function).

As to the rendering problems, they are basically merely rendering problems.  You should be able to change the location of the text as required.

On the Lithuanian product's forum there was discussion in February and March that, by default, the port should be nameless and typeless.  I thought the argument made a lot of sense.

HTH,
Paolo

[davisford]

Hi Paolo -- I read it, but don't remember being clear on the usefulness of it.  I will go back and read it again.

Many times in UML I find that it provides you with aids to illustrate things, but I guess what I am really asking the forum here is if anyone here has found a real practical use for this with what they do.

I have used RoseRT before, and am familiar with the port concept..and other architectural principles like component-connector and port, but to me, when I draw a component diagram, it seems to me that I can say what I want without the port.

Was there recommendation in the Lithuanian forum on whether or not to aggregate multiple interfaces into a single port or not (ala component3)?  I seem to recall Ambler's UML2.0 style says not to do this but I don't remember why -- any rationale or thought on why/why not?

[Paolo F Cantoni]

Hi Paolo -- I read it, but don't remember being clear on the usefulness of it.  I will go back and read it again.

I find I still occasionally go back to the specification.  It tends to make more sense after you have tried some of the things out...  

Many times in UML I find that it provides you with aids to illustrate things, but I guess what I am really asking the forum here is if anyone here has found a real practical use for this with what they do.

I must confess to not having used them myself.  But I think I understand what they are about.

I have used RoseRT before, and am familiar with the port concept..and other architectural principles like component-connector and port, but to me, when I draw a component diagram, it seems to me that I can say what I want without the port.

As has been mentioned many times in these forums, the whole point about modelling is communication!    If you feel you can communicate everything you need without ports then do so...

Was there recommendation in the Lithuanian forum on whether or not to aggregate multiple interfaces into a single port or not (ala component3)?  I seem to recall Ambler's UML2.0 style says not to do this but I don't remember why -- any rationale or thought on why/why not?

From what I recall, no.  I haven't checked out Scott's work in this area, yet.  However, some personal thoughts (from first principles):  One important purpose of the port (I think) is to provide a delegation point for the interface(s).  If more than one interface can be satisfied by the same delegate, then all those that can, should go onto the same port.  Those that can't, should go onto the appropriate port for their delegate.  Make sense?

HTH,
Paolo