- From: Dan Connolly <connolly@w3.org>
- Date: 26 Aug 2002 08:41:02 -0500
- To: Frank Manola <fmanola@mitre.org>
- Cc: Dan Brickley <danbri@w3.org>, Brian McBride <bwm@hplb.hpl.hp.com>, w3c-rdfcore-wg@w3.org
On Fri, 2002-08-23 at 10:28, Frank Manola wrote: > Actually, my original question didn't involve putting words into specs. > However, now that Dan (both of them) have brought up the point, the term > "RDF processor" (or "processor" used in a way that would be taken to > mean that) appears in the current version of the Primer (23 August 2002) > in the following places. If anyone wants to propose wording changes, > I'll be happy to incorporate them. OK, here are some wording changes: strike all mention of 'RDF processor' and 'processor'. Rationale: The technical content of the primer must be a consequence of the technical content of the other specs. None of the specs specify what an "RDF processor" is. > Section 3: > An RDF processor would form the complete URIref of the creation-date > property from the QName <ex:creation-date> by converting the ex: prefix > to the namespace URI defined for it in Line 3, and appending > creation-date to it. Possible re-wording: the complete URI reference of the creation-date property corresponding to the QName ex:creation-date results from replacing the ex: prefix by the namespace URI defined for it on line 3. > Section 3.1: > > This could create a problem, since if an RDF processor retrieved the > catalog from the mirror site, the URIref generated for our example tent > would be http://mirror.example.com/2002/04/products#10245, rather than > http://www.example.com/2002/04/products#10245, and hence apparently a > different tent. Change "RDF processor" to "user agent" or "client" or some such generic term. > Section 4.1: > Note that the RDF/XML uses li as a convenience element to avoid having > to explicitly number each membership property. The RDF processor will > generate the numbered properties rdf:_1, rdf:_2, and so on from the li > elements as necessary. ... each membership property. This syntax is short-hand for numbered properties rdf:_1, rdf:_2, and so on. > The examples above illustrate that the general structures of the RDF > graphs for both Bags and Alternatives are the same (and they are also > the same for Sequences); only the indicated rdf:type is different. RDF > considers anthropomorphism. Please avoid using 'RDF' as the subject of active verbs. > these types as essentially "hints" to a processing application > on how to properly interpret the structures. This is because RDF > processors are not in a position to control how an application actually > uses these structures. For example, an RDF processor has no way to force > an application to use the first member of an Alternative collection as a > default value. Similarly, an RDF processor has no way to force an > application to ignore order in processing a Bag. > > RDF processors are also limited in their ability to enforce structural > constraints on these collections. For example, these structures > explicitly permit duplicate values. RDF does not define a Set container, > which would be a Bag with no duplicates, because RDF processors are not > necessarily in a position to enforce a no-duplicates constraint (for > example, a duplicate might exist somewhere else on the web, unknown to > the processor). Also, if you create the membership properties yourself, > RDF does not insist that the property numbers be contiguous starting > with rdf:_1. For example, you could create a legal Bag with just the > membership properties rdf:_3, rdf:_7, rdf:_8, and rdf:_11 (although an > RDF processor would not generate these property names from a collection > of rdf:li properties). I suggest striking those 3 paragraphs. > Section 5: > > RDF Schema uses the RDF data model itself to define the RDF type system, > by providing a set of pre-defined RDF resources and properties, together > with their meanings, that can be used to define user-specific classes > and properties. These additional RDF Schema resources extend RDF to > include a larger reserved vocabulary with additional meaning. These > resources become part of the RDF model of any description that uses > them, and extend the meaning of that description for any processor that > understands the extended vocabulary. Strike the last sentence. (the phrase 'the RDF model of any description' misuses the term 'model', too.) > Section 5.3: > Moreover, depending on how the processor interprets the property > declarations, an instance might be allowed to exist either without some > of the declared properties (e.g., you might have an instance of ex:Book > without an ex:author property, even if ex:author is declared as having a > domain of ex:Book), or with additional properties (you might create an > instance of ex:Book with a xyz:technicalEditor property, even though you > haven't defined such a property in your particular schema.) > > In other words, RDF Schema declarations are always descriptions of RDF > instance data. They may also be prescriptive (introduce constraints), > but only if a processor interpreting those statements wants to treat > them that way. All RDF Schema does is provide a way of stating this > additional information. Whether this information conflicts with > explicitly specified instance data is up to the processor to determine > and act upon. Strike those two paragraphs. > Section 6.3: > This allows XPackage to be implemented as a general XML application > without an RDF processor, while still maintaining RDF compliance of > conforming documents. Strike it. 'XML application' is no more specified than 'RDF processor'. -- Dan Connolly, W3C http://www.w3.org/People/Connolly/ office: tel:+1-913-491-0501
Received on Monday, 26 August 2002 09:41:06 UTC