- From: Pat Hayes <phayes@ihmc.us>
- Date: Tue, 26 May 2009 18:05:56 -0500
- To: public-rdf-text@w3.org
> > On May 26, 2009, at 3:43 PM, Sandro Hawke wrote: > >> >> Nice explanation. Can you send the RDF' bit, or what you like, to >> the >> mailing list? Which I did, which is why y'all are reading this. "RDF' " is the 'plain typed RDF' in that draft I just posted a link to. >> Thanks! >> >> - s >> >>> On May 25, 2009, at 12:02 AM, Sandro Hawke wrote: >>> >>>> >>>>> Let me try to explain. >>>>> >>>>> I see where Boris is going, but just saying that the literal >>>>> must not >>>>> be used doesn't work. The point is, who has the authority to say >>>>> that >>>>> a certain datatyped literal MUST NOT be used? After all, RDF >>>>> itself >>>>> does not prohibit any legal datatype to be used to type a literal, >>>>> and >>>>> there is nothjng in the idea of a datatype itself that allows a >>>>> datatype to exert any kind of prohibition over RDF syntax. The RDF >>>>> datatype description certainly does not give any provision for a >>>>> datatype to exert any restrictions on RDF syntax. So if someone >>>>> were >>>>> to publish some RDF with literals of the form "foo"^^rdf:text, >>>>> they >>>>> would not be doing anything illegal according to the RDF specs. >>>>> That, >>>>> it seems to me, is the core issue here: a datatype definition, >>>>> speaking precisely, can only define a datatype. It cannot say that >>>>> certain RDF syntactic forms MUST NOT be used. It simply is not the >>>>> kind of spec that can have this kind of authority: that just isn't >>>>> the >>>>> kind of thing that datatypes do. So the second paragraph says >>>>> "this >>>>> specification semantically extends RDF by introducing a datatype >>>>> called rdf:text." but that cannot be right, since just defining a >>>>> datatype cannot itself extend or alter RDF. >>>> >>>> It seems to me that any time people have a reasonable choice about >>>> what >>>> to do with their computers, and it would been good for the world >>>> for >>>> people to act in concert, then it's appropriate for W3C to issue a >>>> Recommendation telling them what to do there. If it were actually >>>> thought to be good for everyone, I think W3C could perfectly well >>>> issue >>>> a Recommendation to never use URL's in RDF which ended in the >>>> letter >>>> "q". >>> >>> Hmm, I guess we think differently. The W3C does not lay down laws: >>> it >>> makes recommendations which define various notations and formats >>> (HTTP, HTML, RDF, etc. etc.) For instance, the W3C has issued the >>> OWL >>> recommendation, but if Im not claiming to process OWL, I can just >>> ignore it. I'm only obliged to follow it if I'm claiming to be >>> conforming to that particular Recommendation's specs. Right now, >>> the >>> use of rdf:PlainLiteral typed literals is legal RDF, according to >>> the >>> RDF specs. I don't see how anything that anyone can say can change >>> that: its just a fact we have to live with. If someone want to say >>> that they are following the RDF spec, and ignoring other W3C specs, >>> then they are perfectly justified in using it, even if some other >>> document says if SHOULD/MUST not be used. They aren't claiming to >>> conform to that other document's views. >>> >>> Seems to me that this is at the heart of the problem here. >>> Defining a >>> datatype is one thing, and imposing a new syntactic condition on RDF >>> syntax is something else. Mashing them is just going to create >>> confusion. (Did I say "going to'?) So, I suggest, we don't try to >>> retrospectively impose things on RDF: we (in effect) define a new >>> RDF' (though that is a VERY small prime, there) on which we do >>> impose >>> our own conditions, and strongly suggest that people use RDF' >>> instead. >>> We point out that unless you are undertaking to utilize this new >>> datatype, then RDF' is indistinguishable from RDF; and even if you >>> are, its still a subset of RDF which can express everything that RDF >>> can in exactly the same way that RDF always has expressed it, so >>> using >>> RDF' is zero cost and can be achieved simply by saying publicly that >>> you are doing it. What it amounts to in practice is, in effect, >>> announcing that you are NOT going to use this new datatype. And >>> then, >>> of course, not actually using it. >>> >>> >
Received on Tuesday, 26 May 2009 23:06:35 UTC