- From: Christian de Sainte Marie <csma@ilog.fr>
- Date: Thu, 23 Feb 2006 20:24:41 +0100
- To: public-rif-wg@w3.org
Dear all, Here are my (quick) comments on the UC. One general comment (already made in different forms by others), is that less UC, but with slightly differently focused, could be better, both in terms of coverage and in terms of impact. Shortly (because I have to go home), I see to broad classes of UC: - UC about porting rules from one platform to another, for different purposes (and thus, maybe, with different requirements on the RIF); - UC about publishing rules for different purposes (and thus, maybe,...). Essentially, UC 3 and 5 fall in the first pot; UC 1, 4, 6 and 8 in the second one. UC 2 might fall in either class, depends on how you interpret it: I tend to see it as a case of porting rules between platforms for the purpose of aggregation, but it could be seen as a case of publishing rules (e.g. by the pharmaceutical companies) for use by, e.g. MEDIC. And 7? Well, 7 is part the first calss, part the second, and maybe part something else and could be in a class of itself :-) Maybe we could consider restructuring the UC section fo the UC&R document along those lines, and probably merge some of the UC (or parts thereof) in the process, hopefully ending with less UC. Another general comment, already made several times, is that the style of the UC should unified. It should be modified, in addition: I seems to me that the point(s) that the different UC purport to make might often be lost because they are not made explicitely enough: either making the important points explicitely in the scenarii (important in terms of requirements or design goals they set or are likely to set on the RIF); or adding systematically a conclusion that summarises those points (or an introduction, although I woudl rather keep the introduction for setting the general scene for the, more specific, scenario). Here below, some UC per UC comments (copied from the comment section in my answer to the poll, for your convenience :-): - 2.1 Information integration (no): This is more a use case for using rules when information is interchanged (to specify/execute the mapping between different information models), than for interchanging rules. The first scenario is at best unconvincing. I do not really understand the second scenario (wrt rule interchange). - 2.2 Decision support (no): Here again, it reads more like an UC for using rules than for exchanging them. The one sentence that suggests rules interchange ("The rules in MEDIC have been automatically generated...") suggest an interesting use case, on the other hand, that ought to be developed. -2.3 Cross-Platform Rule Development and Deployment (yes, but...): An introduction (or a conclusion) should be added, that generalises the scenario and summarising the point(s) this UC is supposed to make (or they should be made more explicit in the scenario itself). This remarks apply to all UC, btw: they should be more uniform. - 2.4 Policy-Based Transaction Authorization and Access Control (yes, but..): The second part of the scenario does not include rule interchange (after the third paragraph). - 2.5 Interchange of Human-oriented Business Rules (yes, but...): This UC should be rewritten to take a more scenario-like form (and shorter). Part of it overlaps with UC nr. 7 (3rd-party rule-interchange services): interchanging rules to achieve a common understanding of, e.g., what a regulation means (how it can/must be implemented; how acceptable compliance can be achieved; ...). However, the publication of regulation, policies, rules etc for compliance enforcing purposes is an important use case (and, indeed, the rules are not necessarily or solely meant for machine consumption). - 2.6 Publication (yes, but...): Style for rule examples should be unified (over the whole section, actually). - 2.7 Third Party Rule-Interchange Services (yes, but...): A bit "heavy" to read... - 2.8 Rich Knowledge Representation (yes, but...): is that really different from UC nr. 6 (publication). - Coverage (no): Should we have "vendor-neutral persistence of rules" as an explicit use case (although is is really a limit application of the cross-platform UC (nr. 3)? The UC do not cover explicitely everything that is important to ILOG (e.g. regarding object model and/or data access, non logical features - actions in the conclusion, etc), although I think they could be modified to. Christian
Received on Thursday, 23 February 2006 19:24:24 UTC