Re: Where are we at? Where to from here?



Re: Where are we at? Where to from here?

From: Pierre-Yves Schobbens <pys_at_info.fundp.ac.be>
Date: Mon, 15 May 2006 10:51:33 +0200
Message-Id: <E2AE164F-FA28-4D97-9BB0-41ADD2B28C61@info.fundp.ac.be>
Le 15-mai-06  01:51, Jrn Guy S a crit :

> Precise UMLeers,
>
> In response to Greg's original post, attached you will find a small  
> figure,
> showing statistics on the growth of UML. If we perceive the UML as a
> (software) system, we will find that its complexity has greatly  
> increased
> over the years.

This is certainly true.

> This affects usability on a theoretical and practical level.
>
This is not clear. I tend to agree, but the additions made to UML  
were done with the aim to increase practical usability.
Are there serious studies on this topic?

> In other words, with a body of specification that large,  
> consistency is
> almost impossible to achieve.

It is clear that the official description of UML (any version)  
contains inconsistencies.

> In the same vein, practicability is even harder, because lack of
> specification forces implementers to interpret the standard, rather  
> than
> adopt it. The result is not UML, but VML - Vendor Modelling Language.

True, but most people don't see this as an obstacle to practical  
usability (I do).
>
> * Abort work on the existing UML standard 2.0.

I disagree: this is the best way to have almost everybody consider  
our work as irrelevant.

> * Focus work on well understood, open, and minimal systems, such as  
> EMOF
> (the basis for IBM EMF). Start with a class-based abstraction, such as
> Object-Z (One option!)

This just about meta-modelling, not about UML itself.

> * Set up a research-based forum of focus, that is independent of  
> the OMG,
> like the OCL Portal
> (http://web.inf.tu-dresden.de/~s9600916/OCL-Portal/index.php)

A forum is nice, but why make it independent of OMG? This will only  
increase confusion.

> * Build minimal metamodels of particular interest for UML's parts,  
> describe
> their conditions of composition into a new metamodel.

Why not use profiles to define restricted parts of UML?

>
> This approach could provide the following advantages/ remedy the  
> following
> shortcomings:
>
> * Research impact is immediate, and the changes made to a commonly
> maintained standard are traceable. The counterexample is the  
> current OMG
> process, were the proceedings of the workgroups cannot be observed.  
> Only
> selected results are published at larger intervals.

This is at it should be -- interested people can have the information  
through the members, while publishing all discussions in extenso will  
rather confuse most people.

>
> * Research is rewarding, because contributions can be ascribed to  
> research
> institutions and individuals. In the current process, researchers  
> may, as in
> the case of the OCL standard, contribute significantly to the work,  
> but do
> not receive a mention in the result. As this is uneconomical for the
> researcher in the promotion of their own work, they withdraw from the
> effort, leaving it weakened.

Well, it is still possible to publish individual contributions, but  
not as endorsed by OMG.
I don't think this is the main problem.

>
> * Results are free. In the current situation, OMG publications are
> copyright-protected content of the same-named company.  
> Consequently, each
> person that wants to use terms branded by the OMG or reference the  
> products,
> has to go through a legal process. This stifles innovation. In one  
> concrete
> case known to me, a researcher wanted to produce a public HTML  
> version of
> the UML standard, to make it more accessible. The request was  
> denied by the
> OMG.

I agree that this is a problem, but not a major one, and the  
situation is the same with almost all standard bodies (CEN, CENELEC,  
ISO, CCITT, etc.)

>
> The last argument I would like to add is of a moral nature: If we  
> do not
> provide a unifying basis to the UML and MOF, there will be no  
> common use as
> explained above. Thus, the biggest shark is going to take the  
> smaller fish
> in the waters of modelling tools. Microsoft has already affected a  
> change in
> the naming of the UML conference from UML to MODELS. It will not be  
> long
> until the UML disappears from public perception and we will have to  
> write
> about 'Software Factories', 'DSLs' and their underlying formalisms.  
> Quoting
> one of Microsoft's Researchers:
>
> "So I don't see UML as a central part of the software factory/DSL  
> story."
> (cmp. http://blogs.msdn.com/stuart_kent/archive/ 
> 2004/09/29/235482.aspx )
>
> I hope that model-driven processes will not become intellectual  
> capital of
> one institution. Digressing from the OMG-standard to avoid such a  
> situation
> seems to be a small price to pay.

It this would be the situation, I would agree. But the fact is that  
in OMG, there is currently a kind of balance of powers that makes it
reassonably fair, though I agree its mechanisms are (in principle)  
unfair.

Anyway, the first point is to prove that we have some value to offer:  
concretely, where would this value be?
Received on Mon 15 May 2006 - 09:52:38 BST