AWWSW -- 04 Mar 2008

<jar> http://esw.w3.org/topic/JonathanRees/Notes

http://esw.w3.org/topic/JonathanRees/Notes

TAG Meeting

<inserted> Scribe: dbooth

<inserted> ScribeNick: dbooth

jar: We discussed the progress of AWWSW at the last TAG meeting.
... Rules really resonated and helped the TAG members. "I was worried about this effort. Now I'm not."

scribe dbooth

<jar> http://esw.w3.org/topic/JonathanRees/Notes

jar: a "potential rep" has the opportunity to be a rep of a resource.
... if an http response has not actually been issued by a resource yet.
... people have complained about the use of rep. So in an ont we need something that might be a rep.

timbl: i'm not too worried about having a separate name for something that could be a rep.
... 2 questions: rel btwn an http response and a rep. does it carry anything else?
... the other q: http spec says it's a rep if the status 200 says its a rep.

<alanr> jonathan calls this a "200 responder"

<alanr> but it is not heard as such in common discussion - there is more to it

jar: 2616 says a rep is an entity that goes with a 200 response, but the entity is not the whole response. it ditches the headers.
... also I was trying to capture the discussion a few weeks ago, someone was worried that restricting to http responses was too limited. wanted a more general description of what a rep might be.
... let's table this and move on to rules.

DBooth Rules

http://esw.w3.org/topic/AwwswDboothsRules

<inserted> Scribe: jar

discussion of jar's table & what's a reprsentation tabled until a later meeting

scribenick jar

the rules are a combination of 3 ontologies - uris, http, web architecture

<alanr> uri:hasProperRacine doesn't work with open world, if the intention is that there is *no* racine for some URIs

<timbl> ######################## AWWW #########################

<alanr> or do these rules assume closed world?

Meant to delete definition of 'resource'

<Stuart> How does awww:Resource differ from rdfs:Resource and/or owl:Thing?

<timbl> rdf:label "InformationResource" should be rdf:label "information resource"

hasRacine, hasProperRacine explained

alanr: are you assuming closed world reasoning?

timbl: the domain statement isn't as constraining as it could be.

dbooth: the correct domain restriction would reflect that the URI on the left of hasProperRacine has a #

alanr: you need a cardinality restriction in order to say what you mean.

timbl: but the logic doesn't capture the entire semantics

<dbooth> dbooth: Right, the domain and range are not as restrictive as they theoretically could be.

<timbl> it never can, so why bother trying to be so experssing in the RDFS.

<alanr> hasProperRacine subProperty hasRacine

alanr: hasProperRacine is a subproperty of hasRacine

<alanr> note for future: Need xsd:URINoFragment

dbooth: hasURI is the inverse of denotes. similar to log:uri

<timbl> Style: "hasURI" is messy, preefr "URI".

<alanr> q: is log:uri inferred by n3?

dbooth: hasDirectGetReply -- grounds this out
... hasGetReply allows for forwarding

alanr: if n3 deduces log:uri relationships then we can just use a type [scribe is confused]

timbl: issues about what has to be asserted or inferred all has to do with your test harness

<alanr> +1

<alanr> the thing that makes sense is to say, the system can not infer x, given y

timbl: request to separate out the two kinds of material in these files

<timbl> Theer is N3 and cwm

alanr: if log:uri has been inferred then there should be no need to assert hasURI

<timbl> N3 allows you to write these rules.

<timbl> Cwm has certian built-ins

<timbl> as well as an understaning of n3

alanr: what is the range of a location? looks like it's a URI

dbooth: it's the location: header value (string)

alanr: can we change the name, for clarity? a string is not a location

timbl: can we use a separate namespace for headers, mapped automatically from header names to ontology URIs?

<dbooth> timbl: I suggest a separate namespace for htttp headers -- hh or httph.

alanr: can I suggest something that strips the "has"?...

timbl: lots of reasons why no "has" is good

alanr: it's known that many people are confused by ambiguity of leaving out the "has"

dbooth: will do both

stuart: no, pick one and stick with it.

<timbl> label "direct reply"

<dbooth> alan: please add rdf:labels to these

<alanr> rdfs:label the other one

alanr: there's no language, wouldn't it make more sense for [...] to be strings?

timbl: mapping things to strings is a whole lot simpler, for testing purposes
... you get a string out of the protocol, then coerce it to a string-that-is-a-URI

<alanr> hasContentType: text/plain@en is not permissible

timbl: xsd:anyURI is awkward and it's not clear it adds anything

alanr: i was asking about has-content-type , which is marked as an rdfs:Literal, similarly status-code
... this would permit things marked with language designators
... if we just say it's a string, we rule out nonsensical cases such as a content-type that's in a language

timbl: timbl has an action item to align tabulator's ontology with this one

alanr: tabulator should use its own namespace

timbl: no, tabulator is trying to be faithful to http

<timbl> "text/plain"

<alanr> "text/plain"^xsd:string

<dbooth> dbooth: if these rules are going to correctly reflect HTTP, then the types should match the http spec s much as possible. I'm not sure what type the http spec says the value of the content-type header is.

alanr: the ^^xsd:string should be inferred if the range of the predicate is specified properly

<timbl> "text/plain"^^xsd:string

timbl: no, doesn't work that way

<Stuart> IIRC typed literals and plain literals are distinct in RDF.

alanr: have to check that one

***** let's plan to leave off at 9:55 so that we can consider next steps (homework) *****

alanr: maybe change the n3 parser to that "string" missing a ^^ is impicitly a xsd:string

<alanr> alan agrees that syntax should match practice

<Stuart> *if* you make that change how would you then enter a plain literal?

ISSUE: use of "..." syntax in relation to strings and literatls -- table

<Stuart> ...and what would you do about all the existing N3/Turtle documents.

<alanr> hasDirectGetReply subproperty hasGetReply

dbooth: next rule is questionable - it says that a 301 means that the new location URI names the same resource as the original
... this is based on the http spec

****** it's 9:55

dbooth: this is a somewhat difficult area
... then what is the purpose of IR if these two things are *not* the same

<alanr> then informationresource not an abstract object?

stuart: which rule?

<timbl> ?r1 = ?r2 .

<dbooth> dbooth: I should give each rule a name, for easier reference

<dbooth> stuart: line numbers would be good.

dbooth: the one whose comment starts "Furthermore, 301, 302, and 307"

<alanr> +1

stuart: suggests everyone read and comment on the rule set, via email, before next meeting

<timbl> http://www.w3.org/2006/gen/ont#sameWorkAs

<alanr> http://www.w3.org/DesignIssues/Generic.html

timbl: different information resources that are the same work - please read the design issues note
... specific vs. generic IRs -- core to the way the web works

<dbooth> dbooth: if the two things should *not* be considered the same IR, then why not?

timbl: timbls's goal for next week: figure same-work-as into this ontology
... you can have uris that aren't "on the web", or that are for IRs but have no reference to the web

[scribe may have made up 'are for IRs' in the previous. need to be careful]

<alanr> permanent redirect shouldn't be from permanent to temporary, though

(discussion about 301, 302, and what they say about where the resource is, now & later)

- DRAFT -

AWWSW

04 Mar 2008

Attendees

Contents

http://esw.w3.org/topic/JonathanRees/Notes

TAG Meeting

DBooth Rules

Summary of Action Items

Scribe.perl diagnostic output