- From: Khalid Belhajjame <Khalid.Belhajjame@cs.man.ac.uk>
- Date: Sat, 16 Jul 2011 14:26:54 +0100
- To: "Myers, Jim" <MYERSJ4@rpi.edu>
- CC: Luc Moreau <l.moreau@ecs.soton.ac.uk>, public-prov-wg@w3.org, renear@illinois.edu
- Message-ID: <4E21919E.40207@cs.man.ac.uk>
> People want to find bobs by their description (not just identifier) – > to me this means our model has to include the idea of properties. If > we punt and say that all means of discovering a particular bob – > finding it based on its description to get to an ID, or using the ID > to learn something about it – is out-of-band, I don’t think we have a > usable standard. If there are alternatives to saying that bobs have > properties with values that keeps this functionality, let’s get them > on the table. (Do any of us think that pushing this functionality > out-of-band is OK?) > Yes, I agree. I actually asked a question in another thread on whether the notion of identity of a BOB is needed, since in general users will be interested in the (content of a) bob itself. Properties is one way of uncovering the content of a BOB, I was, however, more inclined to adopt the notion of "state" since it conveys the same meaning without requiring identifying the properties of a BOB. Khalid > IVPof does have the notion that there is real correspondence between > the two bobs: B IVPof A. I think it is equally valid to talk about > that correspondence in terms of saying that these two bobs overlap in > some of the dimensions in which they are defined, or to say they > represent the same stuff (because there’s something in the overlapping > region of common dimensionality of these bobs). I see the idea of one > bob being a lower dimensional projection of another being the same > special case in which there are properties of A that are mutable that > are fixed in B. Our general case has to include cases where the bobs > are defined in terms of different coordinate systems over those > dimensions (so their properties do not match) as well as where A and B > overlap in some dimensions but each is also defined over some > dimension(s) that the other does not share. (When Allen Renear (U. > Illinois) visited recently, we had a fun discussion of the idea that > only one thing can be in the same place at the same time – dimensional > overlap as a way to define identity/equivalence. He brought up the > example of talking about a statue versus the bronze in a statue as > something that breaks this – I see this as a case where the bronze and > the statue are each defined in dimensions the other is not, which is > why we can’t see one as equivalent to or as a state of the other (not > a hierarchical case)). > > I think what that means in short is that if we try to describe > everything in terms of dimensions, we don’t get something > substantially simpler than we have now. TimBL’s model sounds simpler > because he’s describing a simple subset of what we need to model. > > Jim > > *From:*public-prov-wg-request@w3.org > [mailto:public-prov-wg-request@w3.org] *On Behalf Of *Luc Moreau > *Sent:* Friday, July 15, 2011 7:56 AM > *To:* public-prov-wg@w3.org > *Subject:* Re: simon:entity (or Identifiable) > > Hi Ryan, > > It's exactly where we were about 8 weeks ago. Your proposal seems to > ignore > many agreements this WG has reached. We were using the word "R" then. > > The reason why we moved into the direction of thing and IVP is that > there is stuff out there that is changing. For provenance, we need > something that didn't change from some perspective (or had some fixed > value). > > For your proposal to work from my viewpoint, it needs to be more precise > about what is identifiable. Is it a stuff or is it a state of a stuff? > > I note that all our discussions point to the fact that it is really > hard to distinguish a stuff from its state, since it is very much a > question of perspective. Still, we need to be precise about what is > identifiable. I think that the notion of properties associated with > old:thing/f2f1:BOB is reasonable way of providing the necessary precision. > > Regards, > Luc > > > > Regards, > Luc > > > On 15/07/2011 06:36, Ryan Golden wrote: > > With apologies to Simon for hijacking his namespace, I'd like to take > up Luc's suggestion to break off what he called the "simon:entity" > proposal from the earlier thread into a separate thread. > > Rationale > -------------- > It should come as little surprise that some problems we are trying to > solve by our design have been faced before by others in different > contexts. After poring over the thread between Simon, Jim and others, > I discovered a design issue discussion at > (http://www.w3.org/DesignIssues/Generic), published by TimBL, which > bears a _striking_ resemblance to the discussion we're having on > stuff, thing, entity, entity state, and bob. While he does use the > "R" word in some of the discussion, he makes the key observation that > the identifiers we use every day have "multi-level genericity." That > is to say, some identifiers are very specific ("Halley's comet, as > viewed from the Hubble telescope, on 1/1/2014, in JPG format"), others > more generic ("Halley's comet"). The Web design, he states, "should > not arbitrarily seek to constrain life in general for its own > purposes." Neither should we, I would argue. > > Further, we may may make statements about "dimensions of genericity." > That is to say that a) in relation to the thing it identifies, an > identifier can be generic with respect to a particular dimension, > e.g., in relation to the real Halley's comet, the "Halley's comet" > identifier is generic with respect to time and content-type; and b) > one identified thing may be generic in relation to another identified > thing with respect to zero or more dimensions. TimBL talks about the > relatively small number of dimensions of genericity for electronic > resources, whereas we are interested in the infinite number of > dimensions (i.e., all possible properties) over which identifiers and > things in the world (not just electronic resources) may vary. The > idea of "dimensions of genericity" gives what I believe to be a nice > formulation for what we've been trying to discuss as "IVP of." I > leave the remainder of this discussion to a separate thread, however > (please post any comments on this paragraph to that thread). > > If I fail to express some of TimBL's ideas adequately, I strongly > suggest you read the Design Note--it is brief and more well-written. > > Proposal > ------------- > Given both elegant formulations, I would like to propose we conflate > the following concepts: > old:stuff > old:thing > f2f1:entity > f2f1:bob > f2f1:entity state > > Into a single concept: > simon:entity (alternate suggested name: "Identifiable") > > Which can be described as: > that which an identifier represents > > And, importantly for IVP of: > A simon:entity/Identifiable may exhibit a different level of > genericity in relation to another simon:entity/Identifiable with > respect to zero or more dimensions. > > --Ryan >
Received on Saturday, 16 July 2011 13:27:34 UTC