- From: Pat Hayes <phayes@ihmc.us>
- Date: Wed, 19 Oct 2011 19:03:11 -0500
- To: Sandro Hawke <sandro@w3.org>
- Cc: Andy Seaborne <andy.seaborne@epimorphics.com>, public-rdf-wg@w3.org
On Oct 19, 2011, at 4:14 PM, Sandro Hawke wrote: > On Wed, 2011-10-19 at 14:42 -0500, Pat Hayes wrote: >> On Oct 19, 2011, at 10:28 AM, Sandro Hawke wrote: >> >>> On Wed, 2011-10-19 at 13:32 +0100, Andy Seaborne wrote: >>>> >>>>>>> (1) try to model without using either one, when feasible >>>> >>>> which I took as also pushing RDF lists into the "not preferred >>>> category". That's seems to be letting the technology influence the >>>> modelling too much. >>>> >>>> The fact that an ordering construct is an RDF list which "has issues" >>>> isn't the fault of the modeller. >>> >>> Right... What I was trying to get at here was the advice I often hear >>> that modelers should not be using this kind of construct at all. >> >> Who is giving that advice? And what reasons do they give? OWL uses the vocabulary, so I dont think we should be giving advice which suggests that OWL/RDF is broken or deprecated. >> >>> >>> Honestly, I'm very confused about this bit. Perhaps it's best addressed >>> by having a little bit in the tutorial showing how something can be >>> modeled with lists or without lists, and explaining the tradeoffs. >> >> Im not aware of how to model something that needs lists without using lists. Can you give more details? > > Mostly I'm thinking of something Bijan said, some years back, although > it's been mentioned in recent meetings, too. It's not about modeling > something that needs lists without lists; it's about modeling things > that *don't* need lists using lists. > > That is, Bijan was saying that he's often seen people use lists when > they don't need to. As someone who learned to program using LISP, I find this amusing. There seems to be a presumption underlying all this that lists are to be avoided when possible. Where does this modeling aversion come from? If I could press a lever which caused the whole Semantic Web to convert to LISP overnight, I would do it immediately. > A list of children, from oldest to youngest. A > list of employees of a company. A list of offices, in room-number > order. > > So the point is that folks should consider whether lists are really the > best tool. If they are, then certainly use them. But if the order is > in the data already, or something, then don't use a list. But, at the risk of repeating myself, WHY NOT? > > In that same discussion, he said I shouldn't be using lists to say "and > no more", but should instead be giving cardinality. Another piece of gnomic advice given, apparently, without any justification. > That is, if Alice > wants to say she has two children (Bob and Charlie) and ONLY those two > children, I could use a list: > > :Alice :children ( :Bob :Charlie ). > > but Bijan strongly advocated: > > :Alice :child :Bob, :Charlie > > and some bit of OWL by which I say Alice has two children. It's not > clear to me what's best for this. Ah, I think I see. If you want an OWL reasoner to be aware of the cardinality and use it somehow, then yes, just having a list of two elements does not 'tell' the OWL reasoner anything about cardinality. But that is of interest only in the case where you want to involve OWL in your cardinality reasoning. I think the market for RDF Is much wider than this, and that for other purposes, lists might be just the way to go. > > (I'm also fine with the RDF WG staying out of this modeling topic > entirely.) I would prefer that. If we are going to give advice, let us strive to keep it along the lines if-you-do-this-then-this-will-happen, not we-say-you-should-not-do-this. Pat > > -- Sandro > > > ------------------------------------------------------------ IHMC (850)434 8903 or (650)494 3973 40 South Alcaniz St. (850)202 4416 office Pensacola (850)202 4440 fax FL 32502 (850)291 0667 mobile phayesAT-SIGNihmc.us http://www.ihmc.us/users/phayes
Received on Thursday, 20 October 2011 00:03:43 UTC