Terry,
First of all, I'm thrilled that my favorite writer on the UML actually
replied to Bar Bet #2. I was just at the point of deciding it was a
dud, and trying again to stir some controversy. I'm forwarding the mail
to my mother -- see, Ma, somebody important answered me!
Technically, there is one standard (OMG) answer, and the latest Rose
supports it: we should stop using <<uses>> & <<extends>>, and we should
use <<include>> and <<extend>> with a dependency relationship. The
discussion probably failed to heat up because people have long ago given
up on using these stereotypes "properly", anyway. I *have* used
unidirectional association, at least partly because we haven't upgraded
from Rose 98 yet.
The thing that bugs me most is that unidirectional association with
these stereotypes actually makes more sense to me than dependency.
Clearly, the base use case will not be "compiled" before "linking" with
the extending use case (because we don't compile use cases, at least for
now). On the other hand, association implies messaging, and it seems to
me that, if you can bundle up behaviour into use cases, the bundles can
send messages to each other.
In the end, standards and common usage will trump other considerations.
I accept that. I've been telling people for years that I plan to become
famous by promulgating "Tarkington's Principle", which says: "It's
better that it be standard than that it be good." That's a very
perverse idea to software types, who are nitpicking perfectionists by
nature, but it's amazing how often it applies. Call it "Tarkington's
Perverse Principle", if you like. For me, it applies here.
It's quite clear that dependency is the standard relationship to use
with <<include>> and <<extend>> in use case diagrams, therefore that's
the better answer until standards change. It still doesn't strike me as
a good answer, though.
I'm looking forward to your next version of Visual Modeling with RR,
though I don't expect you to explain why dependency is intuitively the
right relationship in a basic introductory book on the UML.
But, it would be very nice if one of the philosophers or angels at
Rational Corp. has written this up somewhere. Please, somebody, just
give me a sense of the word "dependency" that makes sense in this
context. Bend me, shape me.
-Eric
"Quatrani, Terry" wrote:
>
> Eric,
> Now you can see how hard it is to write a book when things keep changing.
> In the earlier versions of UML (and Rose) the relationships between use
> cases were uses and extends and I think the notation was indeed a
> generalization arrow. In later (and the current) version of UML, the words
> changed to include and extend and the notation was a stereotyped dependency.
> But Rational Rose did not support that (so I used a stereotyped
> association). Now, Rose does support a stereotyped dependency between use
> cases. So the correct answer is "there are two different relationships
> between use cases. They are include and extend. Both relationships are
> stereotyped dependencies. For an include relationship, the dependency
> points from the use case doing the using to the used use case. For an
> extend relationship, the dependency points from the extension use case to
> the base use case (the use case being extended". These relationships will
> be update in the next version of my book (which I am starting to write).
> Hope this helps.
> TQ
>
> Terry Quatrani
> Rose Evangelist
>
> Phone: 610-940-2132
> Cell: 610-659-8272
> Fax: 610-940-2150
>
> -----Original Message-----
> From: Eric D. Tarkington [mailto:[EMAIL PROTECTED]]
> Sent: Thursday, January 25, 2001 2:24 AM
> To: ROSE_FORUM
> Subject: (ROSE) Bar Bet #2
>
> It's been months since Bar Bet #1 (which I really enjoyed, thanks), but
> I guess everyone will forgive, since we are all too busy for humans.
>
> Largely because I'm teaching the UML at Seneca College (School of
> Computer Studies), I keep on asking basic questions over and over. The
> instructors don't agree about some basic stuff, and neither does the
> literature on the UML. We get down to the point where you can't appeal
> to logic and you can't appeal to authority.
>
> So, here's my latest vexed question: In use case diagrams, how do
> association stereotypes know which way to go?
>
> In VISUAL MODELING WITH RATIONAL ROSE AND UML, (c)1998, p. 34, Terry
> Quatrani writes: "Uses and extends relationships must use stereotypes
> because they are both represented by a generalization arrow."
>
> In VISUAL MODELING WITH RATIONAL ROSE 2000 AND UML, (c)2000, p. 34,
> Terry Quatrani writes: "Include and extend relationships must use
> stereotypes since they are both represented by a dependency
> relationship." The rest of chapter 3 contains examples that
> consistently use a unidirectional association relationship, instead.
>
> We're teaching an introduction to the UML using Rose, and we've been
> treating include/extend as identical to uses/extends. So, reading
> Quatrani, we get three options for which relationship "goes with"
> include and extend stereotypes:
> 1. generalization (solid line, closed arrow)
> 2. dependency (dashed line, open arrow)
> 3. unidirectional association (solid line, open arrow)
>
> Quatrani is a genuinely excellent resource, who gives a lot of her kung
> fu along with the facts in her books. Which one of these options is
> wrong? Is more than one of them right? Are uses/extends the same as
> include/extend? When David Duchovny comes back, will there be another
> season of X-Files?
>
> There are actually two bar bets here. My final formulation is:
> 1. Uses/extends is the same as the newer include/extend.
> 2. Unidirectional association is the relationship to which these
> stereotypes should apply.
>
> As before, I won't tell what I think until the pressure becomes
> unbearable.
>
> -Eric
************************************************************************
* Rose Forum is a public venue for ideas and discussions.
* For technical support, visit http://www.rational.com/support
*
* Admin.Subscription Requests: [EMAIL PROTECTED]
* Archive of messages:
http://www.rational.com/products/rose/usergroups/rose_forum.jtmpl
* Other Requests: [EMAIL PROTECTED]
*
* To unsubscribe from the list, please send email
*
* To: [EMAIL PROTECTED]
* Subject:<BLANK>
* Body: unsubscribe rose_forum
*
*************************************************************************
