Re: A few comments on SpecGL -- part 3

All --

Some specific replies to David's comments embedded, as well as some general 
comments about finding our way out of the profiles/modules/levels conundrum...

At 05:27 PM 7/30/02 -0400, David Marston/Cambridge/IBM wrote:
>[...]
>Ck 4.3: I think there is such a thing as a mandatory module,

Or mandadory set of modules.  In the modularized standards that I have been 
working with the most -- XHTML, SMIL, SVG -- there is are concepts of Host 
Language Conformance, Conforming Host Language Document, etc.  To summarize 
and obscure a lot of details -- they rely on a core set of modules which 
must be supported, with additional modules being optional.

Btw, does anyone have other good examples of modularization in W3C, 
different from these three similar ones?  David mentioned XSLT (2.0?).  Any 
others?

>and there
>may well be cases where the implementation of two modules requires that
>a third also be implemented. Whether one module can require another is
>more controversial, since you get into the territory covered by levels.

It may be controversial, but it is done in W3C.  From Ch.2 of SMIL 2.0:

"The modules may be independent or complementary. For example, the 
SyncMaster module requires and builds upon the SyncBehavior module, but the 
PrefetchControl and SkipContentControl modules are independent from each 
other. In addition, some modules require modules from other functional areas."

I apologize to keep coming back to these examples, but they are what I'm 
most familiar with.  I'd welcome any other examples or counter-examples.

This brings me to a suggestion.  If we have well structured and apparently 
carefully considered examples like this, then we shouldn't resist 
them.  I.e., we may eventually idealize a different abstract model for 
modularization.  But here is a model that has been adopted by three W3C 
standards.

In our next published WD, I think that the best we should try for is to 
unify and clearly explain that model, *as well as* any sensible alternate 
models that are proposed.

This same suggestion applies to profiles and levels, and to the 
interactions amongst the three.  For example, I think a lot of the earlier 
models-levels dialog,

http://lists.w3.org/Archives/Public/www-qa/2002Jul/0018.html

contains a lot of explanatory material that could be lifted fairly 
directly.  As far as the 3 combinations are concerned -- how each of 
modules, profiles, levels relates to the other two -- why not just discuss 
the nature of the relationships for now.  Also it is easy to note where 
things are commonly seen (e.g., the XHTML modularization/profiles model), 
or are particularly nicely structured.  For odd or suspect stuff, we can 
(for this draft) somewhat marginalize it, but not try to rule it out.

Thoughts?

>How about this verbiage? "If modules are chosen, indicate any mandatory
>modules, dependencies among modules, or constraints against combined
>implementation of particular pairs of modules." I'm not entirely sold
>on this myself, since I think modules should be, well, modular.

I'm thinking to take your suggested words and put them as bullet-list 
requirements under the tersely worded checkpoint.

>But I
>could also see modules as the most flexible and universal way to divide
>specifications.

That is occurring to me also, as this goes on.

(And levels seem more and more to be a natural consequence of a typical 
pattern of historical development, both in W3C and other venues like 
ISO.  "Levels are a consequence of modularization of time (development).")

More about levels in next (final) message of this series.

-Lofton.

Received on Thursday, 1 August 2002 20:25:32 UTC