Re: DoV and verifiability

On Fri, 6 Sep 2002, David Marston/Cambridge/IBM wrote:

> I think the SpecGL can and does enumerate all the dimensions of
> variability that WGs are allowed to use. It requires the spec to
> convey the outcomes of explicit WG decisions on each one. In turn,
> that means that the WG had to consider whatever variability they
> intend to allow and express it using the conceptual model of the
> eight DoV. This is possible because one dimension, modules, is
> highly generic and can mean just about anything.

You assert that eight DoV is enough for all future specs. We cannot be
sure of that today, of course. It is impossible to tell whether
current DoV terminology/approach will cripple some future specs.

Moreover, you say that "modules can mean just about anything". That
tells me that we may be building a complex terminology/model just for
the sake of introducing something new/neat into SpecGL. That is, we
are developing a model and then forcing anything that does not really
fit into the model by calling it a "module". If module can be
anything, there is no need for the entire DoV model. "X" in the
"minimal set" requirement can mean just about anything.

I would prefer a simple and future-safe scheme:

	- give spec authors full freedom as far as variability is
	  concerned; they should decide what is best for a given
	  spec; we cannot predict the future so future authors
	  get full control; do not tell them how to write a spec

	- check author's work using simple but powerful requirements
	  of SpecGL; these must be universal and simple laws of
	  specs that are unlikely to change (but could be expanded
	  in the future)

	- supply optional tutorials/whatever to teach authors about
	  DoV problems of the past

Even in the simple and straightforward form I advocate, SpecGL will be
immediately useful. As we learn more about SpecGL applications, we
might come back and write SpecGL 2.0 with more complex/specific
requirements, including those derived from the DoV model.


                            | HTTP performance - Web Polygraph benchmark | HTTP compliance+ - Co-Advisor test suite
                            | all of the above - PolyBox appliance

Received on Friday, 6 September 2002 16:48:38 UTC