Proposal to Revise the QAF Document Family
This is a summary of discussions by QA Chairs, who include Karl Dubost,
Lofton Henderson, Lynne Rosenthal, and Olivier Thereaux, plus QA Activity
Lead Dominique Hazaël-Massieux.
Introduction
QA Chairs have looked at recent implementation feedback about the QA
Framework (QAF), and believe that some changes could make it more
user-friendly, more helpful, more welcomed by the Working Groups, and
perceived as less burdensome. A wealth of good material has been developed,
and these benefits can probably be achieved mostly with reorientation and
reorganization.
Rationale
The QA Chairs have taken a look at the status, progress, and W3C response
to the QA Framework (QAF) document family. Based on
- recent W3C input during the CR period of Operational Guidelines (OpsGL)
and Specification Guidelines (SpecGL),
- resources available to the QA Working Group (QAWG) for development of
the QAF,
it becomes clear that some changes are needed to the content, overall
organization, and future development plan for the QAF.
Specifically, we in QA must adjust to two main problems:
- while some QA-friendly groups have received the current QAF positively,
there is also (probably more) indifference and even hostility to the QAF
in its current form;
- and, the resources that QAWG has available to apply to the QAF are well
short of what would be needed to push ahead with the current plan for the
QAF.
This proposal is driven by these two big issues, and some smaller ones as
well. There is a general consensus (amongst qa-chairs) about needed
high-level changes to the QAF family. Lots of specific details are still
undefined, especially about the size and organization of the individual
pieces of the QAF puzzle, and how they are developed.
References
The current QAF family
The current QAF family consists of an introduction and three normative
components, each of which have associated informative components:
(Note. OpsET = QA Framework: Operational Examples &
Techniques; SpecET = QA Framework: Operational Examples &
Techniques)
Major QAF issues
The still-growing QAF CR issues list is at:
The specific conceptual issues that potentially affect the scope &
goals, content, architecture, look & feel, and future processing of the
QAF are: CR-1
through CR-6; CR35, CR-37 through CR-40.
QAF proposed changes
Changes to the QAF components
Proposal for new-QAF:
- Discontinue the OpsGL components (OpsGL, OpsET, and associated
templates) as a normative part of the QAF family.
- Combine the best parts of QAF-intro and the OpsGL/ET components
(including helpful templates) into a user-friendly, non-normative "QA
Handbook".
- Consider changing the language/guideline/document model of
surviving OpsGL guidance along the lines of WCAG20 (see
also current WCAG20 internal draft), or Web
Architecture, or ...
- Advocate for some minimal Charter-commitments requirements about
test materials to be added to W3C Process document.
- Continue to progress a much-simplified SpecGL/ET as a set of normative
Guidelines, with potential changes such as:
- reduce the number of CPs and/or GLs (e.g., perhaps remove P2 and P3
and publish them separately, for now, as a companion "Spec.
Additional Recommended Practices").
- Keep (probably) the DoV concept as an organizing model, but clarify
it and simplify its implementation-related details.
- Concurrently, fully develop the appropriate SpecET parts, including
templates, tools, implementation aids, etc.
- Solve all remaining relevant open SpecGL/ET issues.
- TestGL/ET: progress in background for near future, commensurate with
resources (people) who are interested in and willing to work on it.
Guture coordinated publication with new-QAF as a goal, or not? Evaluate
future coordination with other bits after some trial period with this
approach.
Open questions
- The above proposal defines "what?". I.e., what will be in the final
assemblage that we call QAF. It doesn't define "how?", and doesn't define
any low-level or detailed structure. There are at least two interesting
possibilities:
- new-QAF is a collection of traditional documents, similar to the
current ones; but they are simplified, streamlined, better
inter-related, and completed.
- Karl has suggested an intriguing alternate model under the title
"QAF to FAQ", with several interesting features:
- don't try to maintain our traditional 3-component partition,
because the inter-dependencies end up complicating things too
much;
- instead develop a larger number of smaller pieces... each one
is like a Q in a FAQ, and addresses a single topic. The sum total
of all of 'em spans all of our current most important QAF topics.
Each topic is developed completely with guidelines and
checkpoints (if appropriate), tutorial, techniques, tools,
examples, possible markup and validators, etc.
- some example topics: (spec) scope definition; classes of
products; extensions; test assertions; test case; charter;
conformance section; etc...
- this develop model could be tackled progressively, one Q/topic
at a time.
- In either of those development models, there is the question of how
much coordinated publication we need, versus how much we might be able to
sequence the pieces.
- There are questions of language/style of th epieces. RFC2119-based
commandments ("blah MUST blah")? Or imperative-voice language? Or test
assertions? The answer may be different for different pieces, e.g., the
QA Handbook might use gentler language, and SpecGL might use RFC2119
language.
- The above-proposed lack of coordination between TestGL and other bits
-- does this run afoul of the big issue of JC (all parts must advance
synchronously)? Or does the issue substantially go away with the proposed changes above (esp. #1 - #3).
- ...there are more, but out of time for now...
Schedule considerations
Without any changes to QAF structure/content/organization, we were
guess-timating that a next coordinated publication -- a 2nd Last Call --
could happen in Oct-Nov 2004 time frame. This guess was based on working on
SpecGL/ET first, then OpsGL/ET, and concurrently bringing TestGL to a
publishable 1st Last Call.
For the proposed QAF revision, there are too many open questions to
estimate schedule. However, it certainly (we believe) would not require
more time and effort. Almost certainly, it should be achievable with
less time and effort. Therefore Oct-Nov might be a generous upper limit for
new-QAF: 2nd Last Call of remaining normative SpecGL/ET bits; the QAF
Handbook; ??TestGL 1st LC??.