See also: IRC log
<scribe> Scribe: dbooth
Alan: Sometimes interesting conversations on SemWeb on the ontolog forum. Some discussions of time now. Also discussed "how do i say what an resource is?"
<jar> alan asks if others have read the 'clarifying http' paper
jar: A couple of things bothered me. NOt sure yet what triple.
Stuart: I also posted a response.
dbooth: Stuart provides a prose
rationale for the contradiction; I provided mechanical.
... Note that this triple is necessary for remembering what URI was used:
<jar> (discussion is about the first rule in dbooth's email.)
<jar> 2nd rule says to assert what we learned by looking at the response.
<jar> Use of id= in the HTML implies that the #id URI denotes a portion of a document
<jar> But DocumentPortion is disjoint with foaf:Person. QED
<alanr> that's what owl:imports is for
<alanr> sumo, bfo, dolce
<jar> To get disjointness, probably we should fit our classes into an upper ontology.
jar: Any pushback on David's
... Interesting that there is only one mention of resource ?r in David's analysis.
... Only if you have an explicit resource do you know anything about what the URI denotes.
Noah: What would you expect to learn?
Stuart: In rule 1 he generates ?rdfa and ?f, you learn about that resource from asserting ?f.
Alan: Example suggests another warning: You follow the hash rule and you expect that there will be some statement abot the resource, and if there isn't there's a surprise.
<jar> it seems to be just a coincidence that you learned about a#b by getting a rep for a
Noah: That's been one of my puzzlements.
Stuart: I'd prefer that at this stage we restrict our attention to a monotype -- not RDFa which combines two types.
Noah: So imagine I get a 200 from
a resource and you get back RDF/XML or N3.
... What do we say about what needs to be in that RDF to be appropriate? If I know I want that URI to be a stock quote, but if the RDF says it's a picture of a dog, then it seems like something funny is happening.
<jar> Noah: What should we be able to expect from the RDF that we get from the resource named by the racine?
Noah: If you take XML, the root
of the doc says "here's what i am". So if the root says "I'm a
stock quote" then you're okay, but if it says "I'm a resume",
then it seems bogus.
... One thing about RDF that has puzzled me is that it doesn't come out and say that.
... RDF tells me extra things taht I didn't ask. Is there a clean story that everybody understands?
<alanr> no, and there should be
<alanr> but I think you are missing something in saying that the 302 is something separable.
jar: Probably not. I think there's an understanding that the associated document should tell you info that will be useful to know.
<alanr> The test of whether you got a stock can't be tested with an information.
<alanr> You only know that it isn't a stock quote by making an association between the IR and the thing out in the world
Noah: Let's look at the simple 200 case (not 302 or 303 yet).
<jar> JAR proposes 2nd use case: simple 200 situation (no RDF in the representations)
<alanr> If you know it is an IR, all you know is that it should validate as the mime type
Stuart: At one level, an RDF doc is a description. It describes potentially many things. It may or may not describe itself also. You could say the same thing in the XMl case.
<alanr> it is information "about" something
Noah: I would claim that a stock
quote is info. To assign a URI to it, and return HTML or
text/plain, nobody complains of violating webarch. I probably
cannot return RDF for it. Because if I did that, I would be
saying that the resource is not the stock price, but an RDF doc
with one or more triples in it.
... So I cannot return RDF for an info resource.
Stuart: So what you really has is
an HTML doc describing the stock price.
... You cannot (without much care) con-neg between HTML and RDF. You have to be very careful to avoid having IDs that are both documents and people. In Alan's example, there's a line that gives a hint that the author was confused.
<alanr> this particular case is complicated by the "hash rule". No hash rule -> no problem of this sort
Noah: So the resource either is the IBM stock price, or it is an RDF doc that happens to give you the stock price. You wouldn't conneg between the RDF and the HTML stock price.
Alan: This is a basic problem with info resources. If you get back an info resource, you realy don't know much about it. Just the mime type and a bunch of bits. But a stock price is rooted in something real.
<alanr> "can I convey the stock price to you in bits"?
<jar> noah: not clear where the info that's core to the resource stops and where the fluff specific to the mime type starts
Noah: First, I think it's very important to get this straight before we go on to the more complex cases. In the case of an IR, can I convey the stock price in bits? Yes, so I can give a 200 for it. It has to be in some MIME type, so it sort of follows that there is some ambiguity about where the information stops and the fluff that came with the MIME type started. If we were going to worry about that, we wouldn't allow conneg.
<jar> dbooth: trying to restate what noah has been saying. not clear why there's an ambiguity
<jar> noah: 1. stock price only 2. html document saying "the stock price is..." + the stock price
Noah: if I had two resources w two URIs, one with the name of the company and price, the other a specific HTML doc with stock price $25. In the first case it could use conneg.
<alanr> how do I get the stock price in euros? In yen? Also content negotiation?
Noah: But for the second one, the
rep may be identical, but the HTML is not accidental. It is
supposed to be a fixed representation.
... We need to be answer the question: why can there be conneg?
<alanr> content negotiation is a bad idea. I'm on the record saying this.
<alanr> there's an open tag issue, IIRC saying same
<alanr> I agree with Noah that this question should be asked.
<alanr> what a coincidence, so does the real stock price vary in time
Stuart: Every w3c doc has a current version URI and a "this" version URI, and for a while they are the same. So the resource varies over time. Coherent w how Fielding describes it in REST.
jar: People use the same URI to talk about the time varying behavior and its currernt rep.
<alanr> only useful if what fielding says about REST is coherent
Stuart: Representations do not have URIs.
jar: That slightly contradicts RFC2616.
Alan: How do I get a resource?
<jar> How can I determined (via audit or any other means) that two URIs denote the same resource?
Noah: Each time someome requests a resource, you serve a rep.
<alanr> every representation is a resource
Alan: Can I say that every rep is a resource?
Noah: No, my cynical understanding is that every important resource shoudl have a URI, but reps don't.
Alan: Isn't this saying that anything can be a resource? So how can we say that reps cannot be reps?
<alanr> a representation can be "served as a resource"
<jar> Can I name a representation with a URI? Consensus: yes. (Noah is saying this now)
<jar> Is a representation an IR? Tim says no.
<alanr> Does that mean resources are only things that have uris?
Noah: Every time a rep is served, i assign a new URI to that rep. A get on that URI will tell you what rep was returned at that time.
<jar> A resource is anything that can be named by a URI. (AWWW)
<jar> Time to wind down, everyone
Noah: Stock prices do not in general have URIs by virtue of being served as reps.
Alan: Two things seem to be going hand in hand that should be separated: something is a resource then it has a URI?
<alanr> what about bnodes?
Noah: You could say that resources exist, and can have URIs if you assign them.
<alanr> new class "resources that have URIS". Question is this equivalentclass "resources"
jar: I'm hearing confusion about
resources, web resources, etc. I suggest some careful
... Other suggestions?
<jar> Noah used the term "web resource" ... jar asks what does that mean?
<alanr> I totally agree!
Noah: I tried to lead us toward a simpler direction first, before we hit the more complex case.
<alanr> Only jonathan does ;-)
Stuart: I think we should ground out Alan's PDF file example.
<alanr> "promote something to a resource"
+1 to starting simple.
<jar> Use case: One copies a PDF file to a different server. What are the relationships among the original and copy resources and representations? Alan will expand this
<scribe> ACTION: Alan to post PDF file use case [recorded in http://www.w3.org/2008/01/22-awwsw-minutes.html#action01]
<trackbot-ng> Created ACTION-2 - Post PDF file use case [on Alan Ruttenberg - due 2008-01-29].
<jar> ACTION: jar to reach out to Tim and Pat [recorded in http://www.w3.org/2008/01/22-awwsw-minutes.html#action02]
<trackbot-ng> Sorry, couldn't find user - jar
<scribe> ACTION: dbooth to write up simple case of serving RDF with 200 response [recorded in http://www.w3.org/2008/01/22-awwsw-minutes.html#action03]
<trackbot-ng> Created ACTION-3 - Write up simple case of serving RDF with 200 response [on David Booth - due 2008-01-29].
This is scribe.perl Revision: 1.133 of Date: 2008/01/18 18:48:51 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) Found Scribe: dbooth Inferring ScribeNick: dbooth Default Present: DBooth, +1.617.452.aaaa, Noah_Mendelsohn, Alan_Ruttenberg, Stuart Present: Noah Stuart Jonathan DBooth Alan Got date from IRC log name: 22 Jan 2008 Guessing minutes URL: http://www.w3.org/2008/01/22-awwsw-minutes.html People with action items: alan dbooth jar[End of scribe.perl diagnostic output]