Re: UML-Conformity

[larsrc raeder dk]

> Am Donnerstag, 17. Juni 2004 03:02 schrieb Andrew Ross:
>> On Thu, 2004-06-17 at 04:01, RittervomNie web de wrote:
>> > To enhance the standard UML components a class should have a checkbox
>> > "Interface" to produce a dashed outline. It is not easy to create a
>> > class-like shape containing methods and attributes, stereotypes and
>> that.
>>
>> What version of UML are you basing this on? In the current version of
>> UML (1.5)  interfaces are indicated using the stereotype <<interface>>.
>
> M. Fowler, K. Scott, Addison Wesley, "UML Konzentriert" (I only know the
> german title) and Bernd Oestereich, Oldenbourg "Objektorientierte
> Softwareentwicklung" refer to 1.whatever UML and prefer a dashed outline
> of
> interfaces. If you haver other stereotypes it's a bit irritating to find
> "interface" among them. It's only an enhancement, optional, some tools do
> it,
> I did it the past few years, looks a bit cleaner for people ignoring
> stereotypes.

I'm afraid the official standard (v 1.5) is against you, see
http://www.omg.org/docs/formal/03-03-10.pdf, chapter 3.29.  That said, I
don't think we've been striving to keep compliance with the specs
generally.

>>
>> It's a bit early to be implementing UML 2.0, isn't it? After all, the
>> spec isn't even complete!
>
> UML 2.0 has a lot of new features which will never be used (IMHO). Who
> draws
> exceptions breaking threaded sequences and so on? I've needed Class-,
> Sequence and Activity Diagrams, Use Case only for sketching. Diagrams must
> be
> easy to understand, not fancy 3d rendered piles of crap. Who can insert a
> class diagram containing 25 classes or interfaces with attributes and
> methods
> into a normal PDF? (printable on A4, shippable as book, not as map *g*)

I agree that there's no reason to implement UML 2.0 just because it's
there.  If there's stuff in it that people actually use, then by all
means.  But let's not bloat Dia with things that nobody uses.

>> I wouldn't switch it off, since that would no doubt result in a bug
>> filed to have it added as a feature, even though it's already there
>> (although it could do with some improvement). There is already a bug or
>> two filed about text placement (roles, association names, and
>> multiplicities) for UML associations and messages:
>>

The thing to do (and I'm mentioning this in bug 65430) is to have a
default placement but also a handle so the user can adjust it.  Such a
handle is used in several other objects.  The switch between automatically
placed and user-placed should be controlled in the same way as
auto-routing for zig-zag lines.  Good little project for somebody with a
bit of time on their hands.

>> http://bugs.gnome.org/show_bug.cgi?id=65430
>> http://bugs.gnome.org/show_bug.cgi?id=118313
>>
>> > If these little things would be done by releasing 1.0 or 0.94 or
>> > whatever, including a good documentation, 80 software developers would
>> > use this tool at work. I love it, it's so independent and absolutely
>> slim
>> > designed, 3 already infected...(discoverd it by updating my SuSE Linux
>> to
>> > 9.1)

Can't guarantee anything, but quite willing to take in patches.  0.94 will
probably not contain much more than is in current CVS HEAD.  1.0 has a set
of goals mentioned on the Dia TWiki.

>> I'd be willing to try and convince our uni to ditch Visio (which none of
>> the staff know how to use anyway).
>
> I hate Rational (no interaction in already drawn sequence diagrams, only
> delete and redraw), dislike Visio (too much) and had to use Together. The
> best thing I ever saw was some nice Java-Tool, Composum. There you don't
> work
> diagram-based but document based. You can draw diagrams and insert them
> into
> the document editor. Best feature was to include different diagram figures
> into a sub-diagram and extract these sub-diagrams from the main diagram
> (only
> a thin outline, rubberband, marks included objects, looks like a package).
> Makes it easy to divide a digram into logical sub parts.

An oft-wished for feature:  Heirarchical diagrams.  Suggestions welcome.

-Lars