- From: Sandro Hawke <sandro@w3.org>
- Date: Sun, 24 May 2009 07:36:24 -0400
- To: Bijan Parsia <bparsia@cs.manchester.ac.uk>
- cc: "Peter F.Patel-Schneider" <pfps@research.bell-labs.com>, public-owl-wg@w3.org
> I'm not getting involved....I'm not getting involved...I'm not get oh > who the hell am I kidding?! Part of the problem with arguing about this on public-owl-wg is that you don't have the other side here. Some of us could try to represent their perspective, but that introduces some errors, and if you do convince us, it still hasn't solved the problem. We got the key parties to subscribe to public-rdf-text so we could talk about it directly. That said, I guess I'll explain the interoperability problem as I see it in my RDF work. But solving the problem for me may well not solve it for the commenter here, the SPARQL WG. Quite simply: when I write programs which take RDF data as input, I have to look for the relevant types of literals. If my program is a FOAF display tool, then when it's graph matching to find the name to display, it's going to have to accept Plain Literals and xs:string datatype literals for values. I'll have to decide what to do if I get some data where there's an xs:dateTimeStamp as a name. I might reject the data, I might ignore that person entry, I might ignore that name triple, or I might display the lexical representation of that literal. I might or might not give a warning, while doing those latter options. Now, if people start using rdf:text for FOAF name values, my software will behave badly. Which bad thing it will do will depend on that choice I made above, but all of them will be bad. All of them will make my users annoyed at me (the software developer), and I (in turn) will be annoyed with you. Yes, I could change my software, but unless I see some real benefit from this new datatype, I'm only changing it because you threatened me with a stick; in my world, you didn't offer me any carrot. Sure it's a small change, but from where I sit, I don't happen to see the motivation. I might well push back and tell the users to stop using this new, evil datatype. If they're just doing it because it's in OWL 2, I might get very resentful and negative towards OWL 2. If I'm doing my job really well and reviewing all the W3C drafts relevant to my tool, I more or less have a duty to formally object to rdf:text now, before I'm in this position of being very resentful and negative. Yes, this is just "market pressure", but that doesn't actually mean I'll be happy about it, and I might well try to "market pressure" you back, by organizing a boycott of OWL 2, or whatever. And Yes, it's a very, very small change I have to make to my code. One more bug, in the sea of bugs that is software development. So, maybe I'll only object if I'm feeling like a curmugeon. (Now, it turns out that I *personally* don't really fit this model, because I've been using RDF since before language tagged plain literals came along, and they've always bugged me. Like you, I find the rdf:text a more elegant solution, and in service to that, I don't mind changing my software. But, it seems, not everyone feels like that. And people who don't love that elegance, well, they're letting us know....) Where I end up with this story is this: if we had a chance to tell the story right, and nicely ask everyone to make this change, and they understood it and why it's worthwhile, maybe things would be fine. *Maybe* we can refine the draft enough to where it does that job. We obviously didn't do it with the LC draft, judging by how worried about rdf:text some people got, and I'm not sure we can do it with any draft. In general, when you're asking people to change their code, it's good to have it be part of some package that includes some benefit they can see... -- Sandro
Received on Sunday, 24 May 2009 11:36:35 UTC