- From: Pat Hayes <phayes@ai.uwf.edu>
- Date: Thu, 24 Jan 2002 19:45:30 -0600
- To: w3c-rdfcore-wg@w3.org
Hi people. Sorry Ive been out of the loop while this stuff has been going on. I am trying to catch up. Some comments on recent discussions. First, all the Superman discussion is very nice, but largely irrelevant to RDF. Most of this stuff about who knows which name means the same as which other name, de re versus de dicto, etc., only really has any content when the language contains some kind of equality, ie some way to *say* that two different names (of individuals) co-refer. RDF has no way to say this, and isn't likely to ever have it, so why are we even venturing into this philosophical minefield? In fact, I'd suggest that any example that involves referring to anyone's state of belief, any assertion of co-identity (ie of the form this means the same as that, or this name refer to what that name refers to), or any assertions about what anyone says (when taken literally), should just be ruled out of our discussions as irrelevant and probably misleading. Second, it would be nice if claims that reification is of real use to anyone were backed up with examples of how it is used. This isn't meant to be cynicism, more that examining real use cases is likely to be a much better way to cut to the quick, and so the more examples we have in front of us, the better. I want to see *how* it is being used. Third, I am pretty confident that any particular interpretation of reification can be incorporated in the MT without too much trouble. Really. There seems to be a widespread sense that reification is somehow intrinsically un-semantic, or something like that. What will be a problem is the case where there are several incompatible versions of reification and we want them all at the same time. Even that could be done by offering a kind of MT smorgasbord, though I would rather avoid that if there is any way to avoid it. Fourth. Although the contrast that Dan C. takes a stance on is the one that gets the most attention, there are others that require answers. Dan C wants the subject of the reification of [Mary loves Bill] to be a word starting with "M" rather than a lady in love. I agree; but I want to know if that is a word TOKEN or a word TYPE? Does the reification identify a particular node in a particular graph (which is to be distinguished from a textually indistinguishable node in a different graph at a different URL), or is it something more like a word in the sense of a word in a dictionary? That matters a lot to how it can be used. Seems to me that the former - the token interpretation - is what we would need for any application involving provenance or history. Under the hood, the entire web consists of tp's of one kind or another, after all, and it's tokens that arrive down the pipe at a certain time from a certain source and get stored in memory in a certain place, not symbols. Fifth. (Although this is a slightly different topic, it's related. ) After the recent webont hoo-ha about layering onto RDF, and after huge battles with Jim H., I am more firmly convinced than ever that RDF really, really needs some way to live a kind of dual life, to be a (simple but useful) basic, vanilla notation for expressing simple facts, sure; but also be a generic structure-encoding framework on which to construct more complex notations, which can parse, transmit and maybe manipulate those notations even when it cannot 'interpret' them. And right now, RDF really cannot do both these things at once. Reification has been used, and touted, as the way to do both at once, but I think that this is a really bad, bad way to use reification. If this is what it is for, then (1) it cannot also be used for what the M&S says it is for, or any of this stuff we are talking about; and (2) this is a really clunky, awkward, triple-hogging, ugly, wasteful way to do that. So I would propose that we find, or invent, some OTHER way to do that, and leave reification to do other things. And I will suggest a way to do it in another message, just as soon as I can get the time; it is just a proposal to adopt the 'context' idea used in N3 and make it into a language feature. If that is out of charter (I will argue that it should not be) then we ought to sketch it in enough detail to see how it would work, recommend that the next WG do something like it ASAP, and assume that they will and that will solve that problem, and put it aside. That at least would free up reification from this issue. Pat PS More on reification alternatives in a later message. -- --------------------------------------------------------------------- IHMC (850)434 8903 home 40 South Alcaniz St. (850)202 4416 office Pensacola, FL 32501 (850)202 4440 fax phayes@ai.uwf.edu http://www.coginst.uwf.edu/~phayes
Received on Thursday, 24 January 2002 20:45:21 UTC