- From: Bijan Parsia <bparsia@cs.manchester.ac.uk>
- Date: Mon, 9 Mar 2009 19:45:30 +0000
- To: W3C OWL Working Group <public-owl-wg@w3.org>
On 9 Mar 2009, at 19:26, Alan Ruttenberg wrote: > I've followed up with Marijke about the issue of hasKey. The issue for > her is that "key" has a specific meaning in the ER and Conceptual data > modeling worlds and that the common notion is that keys are > functional. Because of the overlap of our user base with that one, she > suggests at least a short note in syntax commenting on the difference > between our keys and what people will be familiar with, and suggests > that one of our user facing documents show how to map keys in this > other sense to our keys (i.e. hasKey + functional properties). I'm lukewarm trending toward cold. In Jim's case there was an *actual* misunderstanding that wasted time. Here there's not. I don't think we should do more than we did for Jim's case. "Please note that OWL Keys are not functional by default." > In support of her view she offers two citations: Oh please. > "The Entity-Relationship Model-Toward a Unified View of Data",by Peter > Pin-Shan Chen, I'm certainly well aware of that paper and teach it regularly. We're all, I hope, aware of the common case of relational theory where keys are 1-1. > that she considers a foundational paper in the area, It is, but foundational papers don't tell us a lot about our audience. > which says, among other things: "Basically, an entity key is a group > of attributes such that the mapping from the entity set to the > corresponding group of value sets is one-to-one." And in relational theory it's about the uniqueness of a tuple. I'm underwhelmed. It also requires the *presence* of keys. Which we can't enforce. I suggest we do the minimal possible. While I certainly didn't do anything like a comprehensive or random survey, the users I talked to when designing EasyKeys didn't care at *all* about functionality, nor did they mention it. When I asked specifically about it, they were happy at the lack of functionality. People have pressed for InverseFunctional Data Properties *as keys*. "IFDP, aka keys" is a common refrain. IF does not mean F. So, I'm pretty much back to no change, or, at most, the sentence above. Cheers, Bijan.
Received on Monday, 9 March 2009 19:46:08 UTC