W3C home > Mailing lists > Public > www-qa@w3.org > May 2002

Re: levels/options considered harmful

From: Dan Connolly <connolly@w3.org>
Date: 16 May 2002 22:36:38 -0500
To: Karl Dubost <karl@w3.org>
Cc: www-qa@w3.org, Jim Hendler <hendler@cs.umd.edu>
Message-Id: <1021606599.2295.125.camel@dirk>
On Thu, 2002-05-16 at 22:13, Karl Dubost wrote:
> Dan, Did I understand your concerns?

I don't think so.

Let me try to be more blunt: the QA guidelines
say "specify flavors of conformance" as if
having multiple flavors of conformance were
an every day, desireable thing.

Not so.

Interoperability is best served if each spec
has *exactly one* level of conformance. Every
level or option is an opportunity for one
party to rely on feature X while another party
does not support it.

You might say "they shouldn't have relied
on feature X; the spec doesn't require that
everybody supports it." But to the extent
we can reduce the burden on spec readers
to discern such subtleties, we should, no?

Multiple levels/flavors of conformance do not
serve interoperability; they are often a result
of *failure* of consensus. "We couldn't find
a set of 7 features that we all agreed were necessary
and sufficient; we got tired of arguing
and settled for two levels; one with 5 features
and one with 10. The market will decide who's
really right." Sigh. The premise of a W3C working
group is that the market is ripe for a
consensus standard, not a multiple-choice-quiz.

In such cases, the guideline that best serves
quality is: when in doubt, leave it out.
Publish a spec with just the 5 features, and
see if the market supports a version-next
with 7 or 10 features.

Publish working drafts that say "do we need
feature X? is it OK if we cut it out?"
But (as a rule/guideline)
don't claim victory (REC) until you've decided one
way or the other.

> At 21:17 -0500 2002-05-16, Dan Connolly wrote:
> >regarding:
> >
> >"Guideline 3. Specify flavors of conformance.  "
> >	-- http://www.w3.org/TR/2002/WD-qaframe-spec-20020515/
> >
> >That guideline is presented as if different
> >flavors of conformance have no downside whatsoever.
> You refer to http://www.w3.org/TR/2002/WD-qaframe-spec-20020515/#b2ab3d133
> There are cases where conformance could be modular. The examples 
> which is given is the conformance clause of User Agent Accessibility 
> Guidelines.
> For example, it doesn't make sense to impose on a braille user agent 
> to support CSS colors, or CSS positionning.

Perhaps not; but I consider that an exceptional case. The WAI
specs should (and do) take extra measures to explain why
it's reasonable to have such a complex test for conformance.

> Another example, The Process Document is managed as a specification 
> and do not have a conformance section 
> http://www.w3.org/Consortium/Process-20010719/process.html

I don't see how that's relevant; the process document isn't
about interoperable technology, but about predictable, fair process.

> The Guideline 3 which has checkpoints Priority 2, means if you decide 
> to conform to QA Level "AA" (should be changed to avoid confusion 
> with WAI), you have to fullfil all checkpoints 3.*, We do not impose 
> the conformance rules, We say depending on the conformance statement 
> you'll choose, you'll have to respect a kind of organisation of 
> Conformance rules in the table of contents, etc.

I suggest you should advise what sort of conformance clause
best serves the interests of quality and interoperability:
very simple, single-choice conformance clauses: implementations
conform or they do not. (or: documents conform or they do not.)

> The Guideline 1 says 
> http://www.w3.org/TR/2002/WD-qaframe-spec-20020515/#b2ab3d117
> Checkpoint 2.1. Include a conformance clause. [Priority 1]
> So you must have a conformance clause at minimum, if you decide to 
> follow QA rules.
> If you want to make it mandatory for all specs...

No, I'm just trying to see that our experience about what makes
a quality spec is reflected in our QA guidelines.

To wit: SGML had lots of optional features. XML has almost none.
XML is much more widely deployed. That's very, very valuable
experience; let's pass it on.

> you can impose in 
> the pubrules that W3C Specs must respect QA Level A, as it has been 
> done for WAI.
> Dan, Did I understand your concerns? or can you make it more precise? Thanks.
> -- 
> Karl Dubost / W3C - Conformance Manager
>            http://www.w3.org/QA/
>       --- Be Strict To Be Cool! ---
Dan Connolly, W3C http://www.w3.org/People/Connolly/
Received on Thursday, 16 May 2002 23:36:41 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:40:29 UTC