- From: Dave Reynolds <der@hplb.hpl.hp.com>
- Date: Tue, 27 Jun 2006 14:26:23 +0100
- To: public-rif-wg@w3.org
Sandro Hawke wrote: > I've combined the wiki pages to make a new editor's draft of UCR > for discussion tomorrow and an expected decision a week from tomorrow. > > http://www.w3.org/2005/rules/wg/ucr/draft-20060626 > > There's a formatting glitch where the sub-headings are not numbered (for > instance, "Compliance Model" should be "4.1.1 Compliance Model") which > I'll fix in place soon, but otherwise I think the document is stable and > ready for review. (If I've missed something that one of the editors is > doing, please speak up.) Thanks for this. Here's my comments at this stage: ** Issues that should be addressed before publication 1. The separation of alignment CSFs into two "alignment with .." buckets with only one linked to wide-scale deployment is not helpful and can be misread in several ways. The easiest solution would be to adopt DavidH's suggestion (see thread started by Peter) of merging these two, for now at a least. 2. As also pointed out by pfps the "motivates" links are rather inconsistent. By having general requirements like "semantic precision" in some use cases and not in others it implies there is no need for semantic precision in those other use cases which is not correct. We either need an agreed set or we should omit the links for this draft. 3. The example rules do not word-wrap, making the document unreadable in printed form. [Sorry to put a trivial formatting thing in this category but it came up on the last draft and it can't be that hard to fix :-)] ** Less critical comments [Though it'd be nice to address some of them.] * Intro o It'd help to have a description in the introduction of how the document is structured and how the use cases help to motivate RIF which we then break down into specific goals and requirements. * Use cases o Use case 2.2 still isn't that convincing for me, though I'm happy to leave it in there. To make that vision work you need not just RIF but an agreed RIF dialect which every buyer/seller software works within (i.e. basically an agreed rule language). At a minimum the "motivates" list should include "limited number of dialects". o The last para of 2.3 seems somewhat out of place and could be dropped. (I remember how we ended up with it in there and I know we went round this loop before but can't remember what the outcome was). o 2.5 and 2.4 overlap considerably. The justification for having both was that in 2.5 the rules may not be executable and may need to be tagged with status information (as described in the first para). However, the rest of the use case doesn't make it clear which rules are executable and makes no reference to such tagging. Either drop this use case or make the example more clearly consistent with the first para. * Goals o The phrasing of the "alignment with semantic web" CSF doesn't work for me but I'm suggesting a merge anyway (see 1 above). * Requirements o Requirements "compliance model" and "default behaviour" should be combined I think. o Similarly, "implementability" and "standard components" could be merged. o Requirement "XML types" needs rephrasing, I think this is supposed to specifically refer to XML Schema datatypes, "information types" is too ambiguous for me. I suggested a phrasing in: http://lists.w3.org/Archives/Public/public-rif-wg/2006Jun/0128.html However, I guess that's a bit long compared to the other entries in section 4. How about: "RIF must support an appropriate set of scalar datatypes and associated operations as defined in XML Schema part 2 and associated specifications" ? or just " ... operations as defined in appropriate W3C specifications" o Missing requirement on external call out and SPARQL (see same message thread quoted above). How about: "RIF should support an extensible mechanism by which rules can consult external "blackbox" information sources or query processors such as SPARQL data sources." ? * RIFRAF o Drop SeS 7.2 as too special a case of Ses9 (isn't that what was decided at the f2f?) o Drop Prag 3.2 - it is hard to read "test cases" as a discriminator dimension for rule systems. ** Trivial nits o s/eight/ten/ (section 2, para 1) o In 2.4 the paras 3 and 8 both seem to be saying the same thing (two integration modalities), suggest dropping 3. o In 2.5 should expand the SBVR reference. o Suggest dropping para 2 of section 3, this reads more like an advert for the methodology and isn't helped by use of the term "project". Dave
Received on Tuesday, 27 June 2006 13:26:51 UTC