Re: profiles or not [was Re: Modules and levels in a specification]

[thread started on www-qa-wg]

Andrew,

It sounds like our thinking on this is closer than our words have so far 
suggested...

At 02:16 PM 8/7/02 +0100, Andrew Thackrah wrote:

>On 2002.08.07 13:52 Lofton Henderson wrote:
>>Two comments:
>>1.) it is pretty explicit in the QA Activity statement, the QAWG Charter, 
>>and QA Framework Introduction that W3C thinks (collectively at least) 
>>that the WGs are responsible for the success of their standard in the 
>>field, not just for inventing superior technologies.
>>2.) I don't believe that publishing a technical spec and leaving the 
>>interoperability (between producers and consumers) to the marketplace 
>>works (at least not reliably).  I've had direct contrary experience, in 
>>which we had to go back after several years, add a "Rules for Profiles", 
>>define a couple of target profiles, and try to deal with a half-decade of 
>>ugly legacy content and implementations.
>
>  I agree the interop is best *not* left to the market. But it doesn't follow
>  therefore that it should be handled by spec authors either. So who?
>  I have in mind a special interest group dedicated to conformance and
>  interop for a particular technology. There are lots of these groups 
> around in
>  the wider world.

I believe that it is W3C's intention that some serious initial quality, 
test suite, and interoperability work is the *responsibility* of each 
Working Group.  That means that the WG has to ensure that it gets done 
somehow.  It doesn't mean that the WG can't partner or collaborate with 
outside entities, or import or point to test suites from outside, etc.

I agree that there is room for a SIG dedicated to interoperability of a 
technology, after the standard is published.  In fact, it's a great 
idea.  Its scope might even include writing further profiles of the 
standard, for newly emergent application communities.

But the WG must do the foundation work properly, and that involves not only 
technical invention.  The WG must pay some attention conformance and 
quality questions when writing the spec.  Obviously the test assertions 
need to be integral with specification development, and we are saying that 
the basic conformance policy and conformance requirements should as 
well.  (Hence ... SpecGL.)

For example, if a technology is an obvious candidate for profiling, then at 
least write the spec so that profiling is easily supported, now and for the 
future -- modularization, or profile pro-formas, or...  If the WG's 
requirements and use cases indicate one or two particular profiles, then 
why not have the WG specify them as well, concurrent with publication?  The 
earlier the better, so that the implementation market doesn't start to 
fragment randomly.

>  In the case of W3C you mention the charter - so we can see that for
>  our specs such a group would be active within a WG. I'd also go so far
>  as to say that any conformance/interop units in a WG would be members
>  of the QA(IG)and/or steered by it's work.

The latter, at least.  That's why we're putting all the effort into the 
Framework -- to provide a resource and save each WG from having to reinvent 
the same QA stuff.

-Lofton.

Received on Wednesday, 7 August 2002 14:35:57 UTC