- From: Sandro Hawke <sandro@w3.org>
- Date: Thu, 19 Jun 2003 09:18:12 -0400
- To: www-webont-wg@w3.org
There is an @@@ in the middle pending resolution of my suggestion from a few hours ago on linking our documents to each other. -- sandro ================================================================ To: karl@w3.org, www-qa@w3.org Subject: Re: QA Review of owl-semantics In-Reply-To: <a05200f02bad5de373280@[10.0.1.3]> References: <1051801689.6599.267.camel@dirk.dm93.org> <a05200f02bad5de373280@[10.0.1.3]> Dear Karl and QA Working Group members, This is a reply from the Web Ontology Working Group (WebOnt) to your comments (a QA review) [1] on our Last Call version of "OWL Semantics and Abstract Syntax" ("S&AS") [3]. Dan Connolly sent a partial reply [2] only to Karl, which left several issues open. This message addresses all issues except checkpoint 13.2 on which the editors are contacting you separately. First, I need to apologize for us not properly reviewing and commenting on your Last Call draft of the "QA Framework: Specification Guidelines" [4] before your deadline. Each point below, addressing your comments on our spec is in a sense a comment on your spec. We hope that even at this late date you find our comments helpful. 1. We have decided to publish our specification in the form of several documents which are expected to have somewhat different audiences. You have reviewed only one of the six and so missed most of the (informative) background and explanatory material, as well as the (normative) test suite and conformance section. S&AS [3], the document you reviewed, stated: This document is designed to be read by those interested in the technical details of OWL. It is not particularly intended for the casual reader, who should probably first read the OWL Guide [OWL Guide]. @@@ It has since been changed (but not yet republished) to include a more prominent reference to the Document Roadmap in "OWL Overview" [5]. We believe many checkpoints not met by S&AS alone are fully met by the OWL specification as a whole. These include (in the order given by your review) 1.1, 2.2, 3.4, 3.5, 10.1, 13.1, 1.4, 3.1, 13.4, 14.1, 14.2, and 2.3. Most of these are addressed in OWL Test Cases [6], which is in Last Call until 27 June. The remaining checkpoints on which you noted non-conformance are discussed below. 2. Regarding checkpoint 9.1 ("Indicate if the specification is extensible."): this is a difficult and perhaps ill-defined issue in our design space, being layered on RDF. Members of our working group have written academic papers on the subject of syntactic vs. semantics extensibility here. Meanwhile, there is, as yet, no experience with vendors or a marketplace to use as a practical guide. To try to craft a statement which would thread the technical difficulties and yet still help shape a hypothetical future market was not seen as a productive use of Working Group time. Personally, I would be happy to discuss definitions of extensibility in the RDF arena with you, but the Working Group considers this issue closed. [7] 3. On checkpoint 13.2 ("Distinguish normative and informative text") the Working Group feels that the style for making this distinction is a matter of editorial discretion, best done with an understanding of a particular document and its audience. The editors of S&AS have agreed to discuss this directly with you. In more general feedback on your specification, there were some opionions within the working group that might be of interest to you: a. Some people find the QA guidelines impossible to apply in a useful way to their work. OWL is specifically designed to be useful into a future of unforseeable software systems. Its design involves techniques (model theory) which make this possible but which do not necessarily mesh well with such QA notions as "conformance requirements". We believe our documents taken as a whole do still meet the QA guidelines, but to some extent they do so as an aside, in OWL Test Cases [6]; the primary specification (S&AS [3]) defines precisely what OWL is without using the language of conformance requirements. In particular, on conformance keywords (checkpoint 13.1), we generally find "MUST is for agents" [8] a compelling argument. A formal language is what it is; only processors should be considered to conform in the MUST/SHOULD/MAY sense. b. There is also an opinion, widely held in the group, that specifications and rationale should be kept separate. The public list archives provide plenty of rationale, sometimes conflicting; to include that debate in the specification would furthur increase its already significant complexity. I'm reminded of ARM [9] and AXML [10]; providing rationale is nice, but is often best done off to the side for a formal language. (Your rationale sections are much more important, given the human context of your spec.) I hope you find these explanations and comments useful. I would be happy to continue discussion of your specifications. Meanwhile, Karl, please let me know (with a cc: public-webont-comments@w3.org) whether this reply is satisfactory in addressing your group concerns about our specifications. -- Sandro Hawke, W3C (Semantic Web Advanced Development) writing on behalf of the Web Ontology Working Group [1] http://lists.w3.org/Archives/Public/public-webont-comments/2003Apr/0064 and http://www.w3.org/QA/2003/04/QA-Rev-owl-semantics-all [2] http://lists.w3.org/Archives/Public/public-webont-comments/2003May/0002 [3] http://www.w3.org/TR/2003/WD-owl-semantics-20030331/ [4] http://www.w3.org/TR/2003/WD-qaframe-spec-20030210/ [5] http://www.w3.org/TR/owl-features/#s1.1 [6] http://www.w3.org/TR/owl-test/ [7] http://lists.w3.org/Archives/Public/www-webont-wg/2003Jun/0170 [8] http://www.w3.org/2001/01/mp23 [9] http://www.amazon.com/exec/obidos/tg/detail/-/0201514591 [10] http://www.xml.com/axml/testaxml.htm
Received on Thursday, 19 June 2003 09:18:14 UTC