- From: Brian McBride <bwm@hplb.hpl.hp.com>
- Date: Mon, 17 Sep 2001 14:37:23 +0100
- To: Jeremy Carroll <jjc@hplb.hpl.hp.com>
- CC: w3c-rdfcore-wg@w3.org
I suppose I think of this a bit differently. The way I see it we are discussing two things here: 1. when two literals are equal from the point of view of the model theory 2. the matching algorithm an application might use when it does a query or other operation on an RDF model The inclusion of f* mixes these up. As I see it, the specs we are writing define equality for the model theory. We are not defining application behaviour. In which case all f* are actually f. That is of course, unless we want to and can persuade Pat to move to a logic which distinguishes between 'should be' and 'must be'. (only joking, Pat :) An application may choose any matching that is appropriate for that application; I'm told that Geordie's can have a good guess at what some Norwegian means, but I don't think we should build that sort of thing into our specs. Brian Jeremy Carroll wrote: >>Jeremy Carroll wrote: >> >>>There was agreement with the treatment of equality proposed. >>> > > Brian McBride wrote: > >>Jeremy, please can you explain why we need f*. Why not just f? >> > > Out of context that could be misinterpreted :-). > > > The argument, which is not bullet-proof, goes like this. > > For general purpose applications our analysis suggests f, this is hence > our recommendation. > > However, we do not wish to tie the hands of application developers who > have specific linguistic needs. Thus the f* can be read as true by an > application developer who may have a strong default to some given > language. > > e.g. a small company developing tools specifically for the US Hispanic > market may have a default language tag value of "es-US" (I think that's > the right one). Then any RDF input which doesn't say it isn't is then > interpreted as "es-US". > > e.g. in the same scenario, the software tool might need to have a small > footprint database and when an RDF file is loaded irrelevant triples are > immediately discarded. These are distinguished by not having language > tag "es-US". This does seem to give real advantage for f* in this > scenario. > > I don't find this argument convincing; but conversely I am not > sufficiently persuaded that it is the wrong way to go to want to > prohibit it. > In many ways my argument that we are talking about the syntactic > equivalence of terms rather than a language processing model rather goes > against this; & the use of xml:lang="es-US" on the rdf:RDF tag addresses > the default language issue. > > > A further point, is that in M&S xml:lang processing is optional. f* is > hence justified by backward compatibility. The * in f* is the difference > between SHOULD and MUST. > > Bill, do you have anything to add here? > > Jeremy > >
Received on Monday, 17 September 2001 09:41:20 UTC