[sc34wg3] New TMCL slides

Lars Marius Garshol larsga at garshol.priv.no
Tue Nov 10 03:15:15 EST 2009

* Patrick Durusau
> For example, I currently have a "works-for" relationship with  
> different types of entities. Or to recall the multidimensional query  
> requirements I posted yesterday, I am on the way to having a  
> "diagnosed-with" relationship with different *types* of issues.
> If I choose to model an association where the type of the other role  
> player changes as one association type (I can hear Newcomb howling  
> in the distance) I don't see how that disadvantages you or anyone  
> else.

It doesn't directly. However, there is no need for you to have the  
association roles change. Instead, you should have different topic  
types for the different types of issue. That would suffice to record  
the information.

If we extend TMCL to make it possible to model this particular  
approach in detail we need to, well, extend TMCL, make it bigger and  
more complicated. Which means that all TMCL tools will be bigger and  
more complicated, and there will consequently be fewer of them.

So this does incur a cost for everyone that is very real.

Where we are with TMCL right now is a phase that every standard goes  
through, which is where the community starts to discover that there is  
something which the standard does not do. At this point there is  
generally a rush to ensure that every single corner case and use case  
is met.

It's funny but somehow everyone wants a small and simple standard, and  
at the same time no one is ever willing to accept that their pet  
feature might perhaps be omitted.

So at the stage we are at now generally turns into a rope-pulling  
contest between the editors, who want a spec small enough that after  
printing it can still be lifted by a single person, and everyone else,  
who wants their particular feature supported.

Graham keeps saying that TMCL is meant to meet the 80/20 case, but it  
will really require a stroke of luck for that to be true once we all  
return home from Leipzig. At that stage we're more likely to be at  

> I suppose part of my reluctance is that I prefer that we enable  
> users to create their own topic maps using such models and  
> constraints as they find useful and necessary, with a minimum of  
> forcing views of "correct" modeling upon them.

My reluctance is to turn TMCL into a paper mountain in the style of  
ISO 29500.

--Lars M.

More information about the sc34wg3 mailing list