WebOnt & JC issues -- 1st look

Lofton Henderson
Latest revision:
26 January 2004 (AM)

Table of Contents

(Caveat. Numbers in square brackets are references to the "References" section. My apologies that I haven't had time yet to make the references more direct (into source documents) and convenient.)


This is a first look at the voluminous issues and comments against QAF, that were submitted by Web Ontology WG (WO), and by Jeremy Carroll (JC) as personal comments. JC participated in the WebOnt comments, but his own are vastly more voluminous and significantly more aggressive.

WO characterizes their comments [5a] as an "official working group position", from which we may deduce that they expect a formal reply from QAWG. JC is explicit in threatening to escalate all the way up through Chairs, AC, and W3C Management if we do not meet his demands ([4] is his big document, [1]-[3] are email threads).


I have read almost all of the words. There are lots of 'em. I have done a little bit of preliminary analysis (and red-lined some comments in printed copy).

The comments

In a quick reckoning, I identified about 127 comment sections, many having multiple addressable points. I'm guessing that there are 200+ individually addressable comments. For reference, this is more than the sum total of individually addressable comments from all legitimate Last Call review.

Based on LC experience, it would take me the better part of a full work week to analyze, organize, cross reference, research, link to previous related issues, and generate preliminary proposals. It would take us a few months (at least) to process these and generate QAWG actions and replies for all of them.

IMO, a careful and fair treatment of all of these issues, comparable to what we did during Last Call issue processing, will invalidate and obligate us to repeat much of the past year's work.

For this initial look, I'm trying to look at the procedural issues, plus the major/critical substantive issues.

Bottom line

At [4c], JC isolates those issues which he insists be formally addressed (or else!).

At [2a], near the end of the message, JC states his bottom-line requirement for handling his procedural objections.

Procedural issues


The JC comments assert that we have grossly violated W3C process requirements.

IMO, some of the claims are patently and demonstrably wrong, and based on JC misunderstanding of what comprises the normative content of QAF.

IMO, at worst, some items have fallen into gray areas of W3C process (i.e., they are debatable), and/or there have been miscommunications and misunderstandings between QAWG and some correspondents.

Controvening W3M decisions

In [1a], JC points to a W3M resolution and claims that we have violated it. The resolution is, more or less, that Recommendations should not attempt to constrain or coerce the WGs, but rather should address technology requirements.

In [1b], Dom and Karl refute. As they correctly point out, the MUST (and other RFC2119) usage in OpsGL describe what a WG must do to conform to OpsGL. Nowhere does it say that a WG must conform to OpsGL. This (according to previous QAWG discussions and resolutions) is a W3C policy issue.

Resolution (proposed): commenters misunderstand. XyzGL (Xyz = Ops or Spec or Test) defines what an object (WG, spec, TM) must do to conform to XyzGL. It does not require that the object conform. Queue an editorial issue to see if this could be made even clearer in QAF.

This is actually a combination of a procedural and substantive issue. WO [5], sections 1.3, 2/1st, 2/3rd, etc.; [4c] 2nd bullet.

Note. Karl also points our (here or elsewhere) that we don't necessarily want to go to Rec, but that we're using the Rec track at least as far as CR, for trial application and feedback.

Objection to not formally addressing comment

At [2a] JC objects that we did not formally address Last Call comments of his before advancing to CR. He points to his long message of last summer, about the May 2003 TestGL WD, in which he had many criticisms. The main thrust of his message was about the waterfall model, and we did in fact make substantial revisions to TestGL in response.

The comment in question was about the previous topic, that one of TestGL's checkpoints is controvening W3C policy ("don't coerce the WGs in Rec-track documents").


JC also mentions here, but does not incorporate "normatively", that he considers Dan Connolly's RFC2119 issue (many weeks after LC review closed) is in the same "not formally addressed" category.

Objection to not listing open issues when advancing to CR

At [3a] begins a thread in which JC claims a violation of this process rule ("list all open issues when advancing to CR"). It is a variation on his premise in the previous procedural issue -- his comments and www-qa dialog about WD TestGL should constitute open issues on CR-advancement of SpecGL and OpsGL.

Miscellaneous collected alleged pubrules and process breaches

[See the last bullet list of [4c].]

Substantive issues

[Just started here. A few are outlined below...]

QAWG has failed to meet its Charter QA goals

QAWG committed to AAA conformance on everything. It has failed that in several respects: no test materials; no test material commitments; no QAPD; no /test/ pages; no QA moderator; missing several OpsGL required points in Charter; etc. Various aspects of this are brought up by JC throughout [4a], and it is highlighted in [4c].

[Specific links to be provided.]

All of QAF must be synchronized

JC insists at several points of [4a] that the many divergent parts of QAF must be advanced together and at the same quality and maturity. This means particularly:

Further, he wants and insists that everything must be in /TR/.

[Specific links to be provided. But see the first section in [4c] after, "In addition I would make three other formal objections."]

Back to 2nd Last Call

According to JC [4a], the magnitude of the changes minimally necessary to satisfy his objections mean everything must go back to Last Call.

[See the second section in [4c] after, "In addition I would make three other formal objections."]

Failure to conform to RFC2119

This is one of the majors of JC's, referenced from [4c]. See specifically:


This is the LC-67 that we argued at length and that led to a couple of the longest QAIG and QAWG email threads on record. Discussed at several telecons also. The QAWG was unanimous in disagreeing with JC's position, and other correspondents on the QAIG/WG threads also disagreed.

There is ample W3C practice to the contrary, including pubrules and the Process Document.

Btw, JC's position on this also mentions ([4a], sec7.21) Dan Connolly's comment, which is basically identical to the Susan Lesch issue LC-67 (and reference to which we folded into LC-67). JC finds DC's position in MUST is for agents compelling. QAWG did not.

Quite simply, there are two camps on this issue. JC (and WO) gives no new argumentation to add to LC-67 and the lengthy email threads.