Experimental IRC log swig-2009-05-27

Available formats: content-negotiated html turtle (see SIOC for the vocabulary)

Back to channel and daily index: content-negotiated html turtle

These logs are provided as an experiment in indexing discussions using IRCHub.py, Irc2RDF.hs, and SIOC.

02:21:37<drobilla>anyone know of a simple geometric coordinate ontology? (I just need to specify x/y coordinates on a canvas)
03:44:16<drewp>drobilla: i'm using http://fabl.net/vocabularies/geometry2d/1.1/
04:51:40<drewp>what's sparql for "?x that are of class1 but not of class2" (or should i model this a different way?)
04:53:58<drewp>i want to mark some of my nodes as test records, so they're always as good as records but they don't show up in "all record" results on the production site
04:54:30<drewp>i tried ":x a :Record, :TestRecord" but i guess i could use ":x a :Record; :forTesting true"
04:57:33<drewp>http://nedbatchelder.com/blog/200507/booleans_suck.html
04:58:43<LeeF>the SPARQL for it is kind of ugly (the new SPARQL WG will hopefully be improving this). it looks like:
04:58:49<LeeF>SELECT ?x {
04:59:06<LeeF> ?x a ex:GoodClass .
04:59:13<LeeF> OPTIONAL {
04:59:20<LeeF> ?y a ex:EvilClass .
04:59:27<LeeF> FILTER(?x = ?y) .
04:59:30<LeeF> }
04:59:35<LeeF> FILTER(!bound(?y))
04:59:37<LeeF>}
04:59:41<drewp>yow
04:59:45<LeeF>yeah, exactly
05:00:24<drewp>well it's all up to me. ":x :forTesting true" seems a bit nicer, but i don't want to put :forTesting false on all the production ones
05:00:34<drewp>so i have a similarly difficult thing to filter for
05:00:44<LeeF>yeah, it's pretty.... painful
05:01:13<LeeF>the new approach will look something like (design still very much in flux)
05:01:15<LeeF>SELECT ?x {
05:01:20<LeeF> ?x a ex:GoodClass .
05:01:28<LeeF> NOT EXISTS { ?x a ex:EvilClass }
05:01:29<LeeF>}
05:01:35<drewp>perhaps i shuold consider it from the one query that's hard: "find all the production resources" .. and just put a :ProdResource on all the prod ones
05:02:02<LeeF>what SPARQL engine are you using, drewp? some of them have extensions that make this less painful
05:02:18<drewp>is NOT EXISTS a single operator, in that design? i.e. there aren't other combos of NOT and EXISTS?
05:02:21<drewp>i'm on sesame
05:03:09<LeeF>(don't think sesame has an extension, but i'm not certain)
05:03:35<drewp>i think i'll go with ":x a :Resource, :ProdResource" since it's totally simple, and if i miss the ProdResource it should show up fast as a bug
05:03:36<LeeF>there are multiple designs - in some of them "NOT EXISTS" is a single operator, in others there's also an EXISTS form for symmetry, and in some the whole thing is part of a FILTER
05:03:45<drewp>as opposed to messing up the test ones and having them appear on the live site
05:03:45<LeeF>totally simple++
05:04:06<drewp>EXISTS { ... } is just the same as ... , right?
05:58:47<mhausenblas>good morning Web of Data
06:46:49<mhausenblas>ouch. ISWC09 abstract deadline is now one week earlier
06:50:34<mhausenblas>LeeF?
06:50:55<mhausenblas>is the in-use abstract deadline now also one week earlier?
06:52:08<mhausenblas>ACTION confused - seems a bit messy (also the email today with the subject '... Call for Papers - Abstracts due on August 8' and then obviously pointing out June 8 ;)
06:58:05<mhausenblas>LeeF__?
08:29:11<mhausenblas>http://web20blog.org/2009/05/26/a-step-toward-data-interoperability-linked-data-databases-and-avoiding-the-security-headache/
08:29:25<mhausenblas>B:| A Step toward Data Interoperability? Linked Data, Databases and avoiding the security headache
08:29:40<mhausenblas>B: blog post by Ken Fischer. interesting
10:50:22<enjayhch>hello
10:50:52<enjayhch>should rapper be able to parse blank nodes in turtle?
10:51:33<enjayhch>like example 6: http://www.w3.org/2007/02/turtle/primer/#example6
10:52:34<shellac>enjayhch: most definitely
10:53:39<shellac>they share an author / editor
10:55:57<shellac>oh, that example contains a typo. are you seeing "The namespace prefix in "exterm:editor" was not declared."
10:57:19<shellac>oh, and another one. trailing '.' after <http://purl.org/net/dajobe/>
11:02:15<shellac>(jena agrees with rapper, btw)
11:22:59<enjayhch>shellac: thanks - I noticed the namespace prefix typo, but not the trailing '.'
11:23:06<enjayhch>ACTION tries again
11:23:24<shellac>I've emailed the editors
11:24:23<enjayhch>yay, working
11:24:34<enjayhch>thanks shellac :)
11:26:48<tobyink>http://buzzword.org.uk/2009/ttldent/spec
11:27:05<tobyink>C:|ttldent - an RDF serialisation for microblogging
11:27:16<tobyink>C: Yes, I don't deny it's a crazy idea.
11:27:34<tobyink>C: Example input - http://identi.ca/notice/4621236
11:27:50<tobyink>C: Output of example - http://buzzword.org.uk/2009/ttldent/implementation.cgi?acct=tobyink&notice=4621236&format=text/turtle
11:29:46<tobyink>C2: Example input - [http://identi.ca/notice/4621236]
11:30:10<tobyink>C3: Output of example - [http://buzzword.org.uk/2009/ttldent/implementation.cgi?acct=tobyink&notice=4621236&format=text/turtle]
11:31:07<mhausenblas>ACTION notes that he doesn't think that this is a crazy idea, tobyink, otherwise the project proposal we recently submitted would also fall into that category ;)
11:31:53<mhausenblas>and, tobyink, others also don't think it's crazy: http://twitterdata.org/
11:31:56<mhausenblas>http://twitterdata.org/
11:32:11<mhausenblas>D:| Twitter Data - a simple, open proposal for embedding data in Twitter messages
11:32:17<mhausenblas>ok? ;)
11:34:22<tobyink>Hmmm... looks interesting.
11:35:16<tobyink>You might notice the Turtle syntax used in ttldent is somewhat different to normal turtle. e.g. it embraces hashtags.
11:36:13<mhausenblas>yes
11:36:47<tobyink>Also there are a few shortcuts available to help keep things under 140 characters.
11:38:37<tobyink>This comes at a cost though - you need to have context information when parsing the message to make sense of it. e.g. <@foo> refers to the foaf:Person who holds account "foo", but you need context to figure out on what server that person holds their account.
11:39:47<mhausenblas>tobyink: you know http://webofdata.wordpress.com/2009/03/19/microblogging-linked-data-in-disguise/ right?
11:40:07<mhausenblas>there are basically two approaches to it:
11:41:00<mhausenblas>IMO you can do it bottom up (that is, sending structured messages, like you do) or top down, that is extracting the data via NLP from the posts
11:41:20<mhausenblas>ACTION BIAB - TTYL, tobyink
11:44:43<tobyink>cygri: You might be interested in http://buzzword.org.uk/2009/ttldent/spec too - it uses prefix.cc as a resolution mechanism for unrecognised CURIE prefixes.
11:50:37<cygri>tobyink: interesting. cheers! reminds me i have to sink some cycles into prefix.cc to fix a few bugs :-)
11:54:45<cygri>#ttl <#me> cyc:temporalRelationshipCommencingNow dbr:Lunch
11:55:28<tobyink>Awesome - I hadn't thought of IRC logs as an application for it.
12:32:22<tobyink>cygri: re prefix.cc changes - feature request: <http://prefix.cc/all-vann.ttl> should return a turtle file listing all your known prefixes in the format: [] vann:preferredNamespaceUri <http://purl.org/dc/terms/> ; vann:preferredNamespacePrefix "dcterms" .
12:35:53<cygri>tobyink, i suppose you're not happy with http://prefix.cc/vann.ttl.plain ?
12:36:23<cygri>oh wait tobyink, i misunderstood
12:37:48<tobyink>It should be like <http://prefix.cc/popular/all.ttl.plain> but using bnodes + vann vocab to describe the prefixes instead of @prefix.
12:39:07<cygri>tobyink, how about this: http://srv.buzzword.org.uk/turtle/prefix.cc/popular/all
12:40:18<tobyink>I hadn't spotted that the pages used RDFa. Now I feel sheepish!
12:42:34<cygri>tobyink, yeah the best place to hide some RDF is IN PLAIN VIEW ;-)
12:43:38<cygri>the rdfa still needs some love, and probably some direct turtle+rdfxml output makes sense... also, the "popular" section is not the most obvious place to look for "all" prefixes...
12:57:32<gromgull>Does anyone know of a tool for doing skos -> graphviz dot files?
13:03:12<tobyink>gromgull: rapper may well be your friend.
13:04:06<tobyink>It can do RDF -> GraphViz (though it doesn't have any specific SKOS support).
13:09:34<gromgull>hmm - but I don't want the RDF graph - i want the SKOS hierarhcy
13:10:59<shellac>trim it a bit (try a CONSTRUCT) then do the rdf -> graphviz thing
13:33:13<Shepard>gromgull: the DOT format isn't too complicated, it didn't take me long to write a serialiser for it for Sesame. shouldn't be too complicated to do this for SKOS
14:10:31<gromgull>Shepard, shellac: I guess a tiny bit of python and rdflib it should be ther
14:13:05<shellac>yep, whatever your comfortable with. dot is very simple
14:13:14<shellac>s/your/you're/
14:15:22<shellac>I nearly tried it in xslt, but realised I have actual work to get on with
14:17:09<mhausenblas>gromgull: will you be at ESWC09?
14:20:15<gromgull>mhausenblas: i will!
14:24:11<mhausenblas>ACTION un-registers
14:24:13<mhausenblas>:)
14:24:39<mhausenblas>nah, just kidding - actually looking forward seeing you (again ;)
14:25:31<gromgull>:)
14:31:07<mhausenblas>hehe, yes this is what can happen if we blindly trust 'the machines':
14:31:08<mhausenblas>http://revealingerrors.com/apostates
14:31:30<mhausenblas>(btw, one of the great sites which bring it to the point)
14:32:04<mhausenblas>E:| Quorum of the Twelve Apostates
14:32:35<mhausenblas>E: spell checker introduced typo with quite some consequences ;)
14:50:51<PovAddict>more like: idiots introduce typo using a spellchecker
14:51:26<PovAddict>the first thing we do is assign blame, and blaming it on the tools is quite common
14:52:42<mhausenblas>PovAddict: if you explore revealingerrors.com a bit, it's not always that easy ;)
14:53:03<PovAddict>the first thing we do is assign blame, and blaming it on the tools is quite commonchat_summary
14:53:30<PovAddict>ugh
14:54:00<PovAddict>I hate IRCing by sshing from my ipod to my computer... mobileterminal sucks
14:54:29<PovAddict>...
14:54:52<PovAddict>yes I just blamed my tools I'm an hypocrite
15:01:46<mhausenblas>hehe
15:08:28<fgiasson>mhausenblas: I would suggest *not to* re-use dcterms:isPartOf to link a resource to its dataset, but to create a Void property of its own
15:09:13<fgiasson>(for void v2?)
15:09:32<gromgull>Behold! Skos2dot http://pastebin.ca/1436341
15:09:45<tobyink>fgiasson: agreed. :England dc:isPartOf :UK , :Europe , :World . :England dc:isPartOf :dodgyDavesListOfCountries ?
15:11:09<fgiasson>tobyink: this is not the real problem I think. It is when you start using isPartOf for different purposes. Then you application has to do much more processing to figure out how the property isPartOf is used in a certain context
15:11:34<fgiasson>for example: you have a bibo:Article which is a part of a bibo:Collection, and a void:Dataset
15:12:02<fgiasson>an application will manage the isPartOf a bibo:Collection differently than with a void:Dataset
15:12:42<fgiasson>so, to figure out what is what, you have to check the type of objects of these properties (additional processing time)
15:13:43<fgiasson>when I think it would make sense that Void describe all its own properties related to dataset (I am not sure it is a good thing to reuse external properties in the usecase of Void). But I would possibly create sub properties relationship with these external properties
15:13:54<tobyink>isIncludedInSet and isPartOf are different sorts of relationship, even if in English the words we use to describe the relationships are similar
15:14:56<fgiasson>where isIncludedInSet come from?
15:15:07<fgiasson>well, it can be void:isPartOf
15:15:19<tobyink>From my head,
15:15:24<fgiasson>the point is *not to* reuse dcterms:isPartOf :)
15:15:37<fgiasson>ok, because it is not currently in the usage guide
15:15:47<tobyink>It wouldn't be.
15:21:09<tobyink>gromgull: Couldn't get your script to was as-is with more recent version of rdflib, but this seems to do the trick - http://pastebin.ca/1436356
15:21:54<gromgull>topyink: yes - it's the svn version of rdflib
15:22:00<gromgull>although I think the recent release is also close
15:22:25<gromgull>but the modules have been moved around
15:25:27<mhausenblas>fgiasson: why?
15:25:47<fgiasson>please re-read the above :)
15:26:21<fgiasson>the example with the bibo:Collection and void:Dataset particularly
15:27:03<mhausenblas>sorry, yeah
15:27:04<fgiasson>(one amongst many others)
15:27:23<mhausenblas>ahm, well, we did a careful analysis and had that discussion
15:27:34<mhausenblas>found that isPartOf is not really used
15:27:42<fgiasson>and what was the outcome of this discussion?
15:27:56<mhausenblas>so we introduced the semantics for it in the realm of datasets
15:28:08<fgiasson>and because it is not yet widely used, you keep dcterms:isPartOf?
15:28:13<mhausenblas>but we can have this discussion again, now with external input
15:28:25<mhausenblas>I'm not insisting on using it
15:28:46<mhausenblas>I just tell you that we didn't blindly guess but we had good reasons ;)
15:29:09<tobyink>fgiasson: I just use "is void:exampleResource of" for every resource. ;-)
15:29:22<fgiasson>well; what I would like to know are these good reasons; so that I find a potential issue in my own reasonning :) (can't think of everything ;) )
15:29:25<mhausenblas>so, as I said, yes, we can create an issue (you want to do this yourself?) and we will discuss it
15:29:59<mhausenblas>oh, I'd have to dig trough the meeting records to precisely give you all the steps
15:30:16<fgiasson>if you have some time, this would be appreciated
15:30:25<mhausenblas>but IIRC we discussed that for quite a while
15:30:33<fgiasson>otherwise, are these records available somewhere? and where do we create issues?
15:30:43<fgiasson>yeah; I would guess so
15:31:04<mhausenblas>fgiasson: records are group-internal, sorry ;)
15:31:09<mhausenblas>issues: http://code.google.com/p/void-impl/issues/list
15:31:28<mhausenblas>please label with 'Release2.0'
15:31:30<fgiasson>good thanks
15:31:56<fgiasson>well; it would be helpful to know what the discussions around Void have been; and motivate the decisions, etc.
15:32:05<mhausenblas>I bring it up next time we have our telecon (which reminds me that a certain slacker still hasn't filled in the darned doodle, right, cygri ?)
15:32:11<fgiasson>(since it appears that there is no mailing list)
15:32:21<mhausenblas>and if everyone is fine with it I'll post, ok?
15:32:30<mhausenblas>there is one (again, group-internal)
15:32:41<mhausenblas>and the public-lod mailinglist
15:32:46<fgiasson>yeah, this is what I am wondering: why internal :)
15:32:48<mhausenblas>where review/input is appreciated
15:33:06<mhausenblas>fgiasson: this helps us focusing, lot of admin stuff, etc.
15:33:31<mhausenblas>it happened to be that there are four people that started this work and well ... there we go
15:33:32<fgiasson>sure; I am not saying doing all the admin stuff there; but maybe to open one publicly
15:33:53<mhausenblas>again, fgiasson the public list we're using is: public-lod@w3.org
15:34:02<mhausenblas>why should we create a new list?
15:34:07<mhausenblas>all the people are there anyway
15:34:12<fgiasson>ok
15:34:17<mhausenblas>this just causes fragmentation, IMHO ;)
15:34:35<mhausenblas>and, as I said, feeback, proposals, etc. are very welcome there
15:34:55<mhausenblas>and we always pick up everything and handle it with care, believe me :)
15:35:02<fgiasson>well, then why not only using one ML for all SW discussion?
15:35:09<fgiasson>sometimes, fragmentation is good ;)
15:35:20<mhausenblas>true. jsut a matter of granularity.
15:35:40<fgiasson>one ML per public ontology development project (I think)
15:36:02<mhausenblas>ok, this is a valid approach. we chose another ;)
15:36:11<mhausenblas>and so far it worked quite well.
15:36:33<mhausenblas>we also have an identi.ca group, #void, if you prefer this :D
15:36:46<mhausenblas>no. seriously. what's the point?
15:36:55<fgiasson>well, I am not that sure for the only reason that I don't have access to the rational that leaded to the reuse of dcterms:isPartOf :D
15:37:27<mhausenblas>on public-lod more people are reading, as we could reach out on any dedicated list
15:37:41<mhausenblas>and I already promised you to take care of it, didn't I
15:37:48<mhausenblas>(in public)
15:37:51<mhausenblas>:P
15:38:15<fgiasson>public or not, I only want to know :)
15:38:35<mhausenblas>sure thing. and we gonna provide you with it, rest assured ;)
15:38:55<mhausenblas>sorry, gotta run for a minute - BRB
15:40:21<fgiasson>good later
15:43:26<mhausenblas>ok, back ;)
15:55:10<kwijibo>fgiasson: , not read whole scroll back yet, but i think voiD might not be making the mistake you think it is making
15:55:53<kwijibo>we don't use dct:isPartOf to link a resource to a dataset that describes it
15:56:17<kwijibo>we use it to link a document describing the resource, to a dataset
15:57:04<kwijibo>eg: <> foaf:primaryTopic <#BigBen> ; dct:isPartOf </meta.n3#Dataset> .
15:58:12<fgiasson>okay; and anything to link a resource to a dataset?
15:58:41<fgiasson>for example: what if I have a file that is an aggregation of resources description from multiple datasets?
15:59:59<kwijibo>fgiasson: (and tobyink) I think there is a problem in linking a resource to a dataset
16:00:29<kwijibo>because (IMO) a resource cannot be part of a dataset - only a description of a resource can
16:01:51<fgiasson>well; if you want to go there.... anyway, my usecase remain:
16:02:06<fgiasson>what if you have one document describing resources from multiple datasets?
16:02:51<tobyink>void:exampleResource links a resource to a dataset though, doesn't it?
16:04:39<kwijibo>tobyink: yes, and i feel kind of funny about it
16:05:20<kwijibo>fgiasson: maybe you could split the document up into sections using sioc or so
16:05:52<kwijibo>sioc:Item dc:isPartOf <meta.n3#voiD-dataset>
16:06:00<kwijibo>sorry
16:06:00<tobyink>If it's OK to have foaf:primaryTopic links from RDF files to non-information-resources, then why not void:exampleResource links?
16:06:01<kwijibo>not that
16:06:21<kwijibo>[ a sioc:Item ; dc:isPartOf <meta.n3#voiD-dataset> . ]
16:06:23<kwijibo>that
16:06:25<fgiasson>kwijibo: sorry but this is a bad answer to me. To me, the document is immaterial. What I want is to keep provenance of resources (whatever they document where they have been serialized)
16:07:16<kwijibo>fgiasson: I think it doesn't feel right to say that a resource is part of a dataset - Big Ben is not part of dbpedia, only a description of it is
16:08:39<kwijibo>fgiasson: I'm not saying my sioc:Item solution is the perfect answer - I just thought of it just now, and indeed, the rdf:type of the document section is immaterial
16:08:51<kwijibo>but why doesn't it answer your use case?
16:09:00<kwijibo>oh
16:09:01<kwijibo>i see
16:09:07<kwijibo>sorry
16:09:09<tobyink>kwijibo: I agree, Big Ben is not part of dbpedia, but it can still be a void:exampleResource. (Big Ben is an example of the kind of resource that dbpedia has data on)
16:09:41<kwijibo>tobyink: yes, so void:exampleResource is not /too/ bad .. you just have to be careful how you use it
16:10:24<kwijibo>but directly saying :RealWorldThing :isPartOf :Dataset
16:10:30<kwijibo>i don't think that makes sense
16:10:40<fgiasson>a resource is a description of a thing; it is not the thing itself, please.... these discussion never lead to anything anyway.
16:10:40<fgiasson>here we are talking about describing knowledge and not any Web Arch stuff should affect any ontology developoment (IMO)
16:10:44<tobyink>Really, there should be a void:ResourceSet which acts as a counterpart of void:Dataset. void:Dataset represents the set of triples in the dataset; void:ResourceSet represents the set of things described by the Dataset.
16:10:56<fgiasson>bbk; lunch
16:11:49<kwijibo>tobyink: that seems like it could be a solution
16:12:32<tobyink>void:ResourceSet rdfs:subClassOf rdf:Bag .
16:13:59<tobyink>{ ?x is rdfs:member of ?rs . ?dataset void:hasResourceSet ?rs . } => { ?dataset void:exampleResource ?x . } .
16:15:11<kwijibo>i think perhaps we need to tighten up how void:exampleResource and void:uriRegexPattern are specified
16:17:08<kwijibo>for example, if my LOD dataset is published at http://kwijibo.dataincubator.org/ but i don't want to, or can't, publishes dc:isPartOf backlinks to the dataset
16:17:59<kwijibo>then void:uriRegexPattern might seem a good way to claim provenance over resources in the http://kwijibo.dataincubator.org/ urlspace
16:20:54<kwijibo>but if I want to use void:uriRegexPattern in another way - to say that there are some descriptions of http://dbpedia.org resources mirrored in my dataset as well, then that use conflicts with the first use
16:21:02<kwijibo>*use case
16:35:37<enjayhch>ACTION heads home
16:58:51<kwijibo>i don't think I agree with fgiasson's statement "a resource is a description of a thing; it is not the thing itself"
17:00:25<fgiasson>kwijibo: please give me one concrete example where your application is made better with all the web httprange-14 discussion?
17:03:05<kwijibo>i just mean, i think a URI does identify a thing, and i thought that's how it worked, i thought that's what everyone had always been saying in all the tutorials etc
17:04:13<kwijibo>otherwise stuff like owl:sameAs would be completely nonsensical
17:05:39<fgiasson>some yes; all I don't think so
17:08:40<Shepard>please not another identity crisis, I think we've had enough of that
17:09:03<PovAddict>lol
17:09:20<PovAddict>httprange-14 took quite a few years to get resolved
17:09:36<PovAddict>and I took a few days to "get" the problem and the solution
17:09:45<PovAddict>so please don't change things around again :P
17:10:38<fgiasson>anyway, httprange-14 has nothing to do with ontology development (which is the point)
17:11:43<kwijibo>fgiasson: it kind of does when httprange-14 is about naming things on the web, and void is about making statements about groups of things named on the web
17:11:52<fgiasson>as Pat Hayes said: don't give meaning to something that doesn't have any (URI)
17:12:07<fgiasson>(except that it is a unique identifier, of course)
17:12:17<kwijibo>I can see how you could argue that an ontology about elephants isn't impacted upon by webarch
17:12:26<fgiasson>why the restriction "on the Web"?
17:12:45<fgiasson>is a dataset only a logical aggregation of things on the Web?
17:13:10<kwijibo>the things don't have to on the web, but the dataset does
17:13:20<fgiasson>couldn't Void be used in a SW oriented desktop application?
17:13:20<kwijibo>the void:Dataset anyway
17:13:33<kwijibo>locally i suppose
17:14:01<fgiasson>ho right; there is this assertion in Void's description
17:14:07<fgiasson>(intro)
17:14:26<kwijibo>but in that case i assume you wouldn't want to publish that you had a private dataset living on your laptop ?
17:14:40<fgiasson>why publishing?
17:15:12<kwijibo>voiD is targeted at publishing descriptions of datasets published on the web
17:15:30<fgiasson>why I couldn't use Void to describe datasets that are used by a local software. So, using void to have some maintenance information that I can use for any purpose (aggregation, statistics, internal application organization, etc)?
17:15:40<fgiasson>yeah, this is what I just read :)
17:15:56<fgiasson>then this lead to a new question: why?
17:15:58<fgiasson>:)
17:16:01<kwijibo>i suppose you could if you wanted - but it wasn't designed with this in mind
17:16:28<kwijibo>because there would have been quite different use cases and requirements
17:17:40<kwijibo>so, i dunno - does that explain why?
17:18:17<kwijibo>basically, it didn't seem important to the authors to incorporate other kinds of dataset into the vocabulary
17:18:44<kwijibo>so we didn't want to claim that it would be useful for anything other than web-based datasets
17:19:40<fgiasson>ok good
17:19:48<kwijibo>in fact we almost made it a lot stricter (only LOD )
17:19:50<kwijibo>gotta go
17:19:53<kwijibo>sorry
17:22:22<drobilla>ACTION wishes turtle was trig
17:25:06<drobilla>turtle and trig should have some kind of in-document identification, ala doctype
17:25:39<drobilla>is this file turtle? trig? n3?
17:26:02<Shepard>they have different mime types...
17:26:38<drobilla>my file system doesn't have mime types ;)
17:27:02<drobilla>on the web it's a non-issue, sure
17:27:14<Shepard>which operating system?
17:27:29<drobilla>lets go with 'any'
17:27:39<drobilla>e.g. fopen doesn't give you a mime type...
17:28:22<drobilla>the perfect world where that kind of meta information is always trucked along with files would be nice, sure, but it's not this one :)
17:28:50<Shepard>well, if you can't figure out the mime type of a file, you're screwed anyway, no? if the file extensions are not helpful enough then yes, you have a problem, as the magic bytes would be the same
17:29:45<fgiasson>why talking about mime types here?
17:30:09<fgiasson>we are talking about describing resources that results in a graph of triples
17:30:19<fgiasson>this has nothing to do with any serialization stuff
17:30:56<Shepard>no it doesn't, I don't think drobilla was relating to that topic :)
17:31:06<drobilla>Shepard: in some kind of abstract theoretical sense, I guess
17:31:18<drobilla>Shepard: but virtually every local/native/etc program ever does this
17:31:45<drobilla>Shepard: using the file extension mostly, but I would like to avoid that (since I want to use an application specific one for that reason)
17:32:20<drobilla>@doctype <turtleuri> would mean the redland guess parser could figure it out no matter what
17:32:32<PovAddict>if they were XML, you'd use the namespace URI :)
17:33:03<drobilla>also useful for stupid text coming over the wire or an fd or something (these syntaxes make a pretty good streaming syntax as it happens)
17:33:11<fgiasson>shepard, drobilla: oups; it was another conversation, sorry ;)
17:33:22<PovAddict>drobilla: your application protocol should tell you what it is
17:33:27<Shepard>if you don't want to create compatibility issues (because the parsers out there don't know this keyword) then you could use a special comment for now. like the hash bang :)
17:33:28<drobilla>fgiasson: yeah, just my own completely unrelated rantings :)
17:33:42<drobilla>PovAddict: true
17:34:19<drobilla>Shepard: the @ directive should be extensible :)
18:26:27<MacTed>parsing "naming things on the web" seems a large part of the httprange-14 comprehension challenge... "naming-things on-the-web" <> "naming things-on-the-web".
18:33:14<kwijibo>MacTed: true
18:42:27<kwijibo>fgiasson: is your objection to voiD's advocated use of dcterms:isPartOf, that it is too generic?
19:34:23<mhausenblas>good nite Web of Data
20:03:42<fgiasson>kwijibo: at first yes; but after the "document, resource, httprange-14, LOD dataset description only, ..." discussion, I found it was deeper than that :)
20:04:46<kwijibo>well, we have to be clear and concise in itemising the issues :)
20:05:47<fgiasson>well; if void's role is only restricted to LOD dataset; then I have do create/use something else; that is it
20:06:13<kwijibo>fgiasson: no, we broadened it slightly beyond LOD
20:06:17<kwijibo>but not much
20:07:05<kwijibo>you are free to use it for whatever you want, if you think it fits, but we are only really keen on developing it for RDF data on the web
20:07:09<fgiasson>I was thinking of it as dataset description... any kind of dataset (with dereferencable URIs or not)
20:07:31<kwijibo>well, the uris don't necessarily have to be dereferencable
20:07:42<fgiasson>and in that case, I found dcterms:isPartOf too general, and I though that it would be good to have a subproperty of this property in void
20:07:54<kwijibo>but voiD is not meant for CSV files or something
20:08:28<fgiasson>is a dataset only a "file"?
20:09:21<fgiasson>to me a dataset is a logical aggregation of data which as a core characteristic which is: provenance; source.
20:10:03<kwijibo>yes
20:10:19<fgiasson>this doesn't necessarly has to do with the way these datasets are delivered (files, API endpoints, dereferencable URIs, etc)
20:10:44<kwijibo>no, it doesn't
20:10:47<fgiasson>but the communication channel(s) can be describe as well for conveniency in some usecases
20:15:58<kwijibo>fgiasson: i would say that's pretty much the case with voiD - we developed it because we felt a need to describe RDF, Web of Data datasets, so we optimise for that, but there is nothing to say that the data must use http uris, or have a sparql endpoint or whatever
20:16:43<kwijibo>however, if you were designing a vocabulary for desktop datasets, it might have quite a different set of classes and properties
20:17:22<fgiasson>okay great; this is what I wanted to hear; however the introduction paragraph would need some re-work then.
20:17:22<fgiasson>if this is the case, then I still suggest to think creating a void:isPartOf or something, property
20:17:42<fgiasson>hoo sure
20:17:59<fgiasson>but this is not my use case; was just to know if it could fit for this :)
20:31:05<kwijibo>fgiasson: no, I don't think we would change our definition of a dataset, I think it is a good one (needs to have a single provider - eg, "the SIOC-osphers" is not one dataset, needs to be RDF, and needs to be accessible somehow - so it doesn't need to have dereferencible uris, or SPARQL, but the dataset description should tell clients how to get at the data, otherwise, what is the point ?)
20:31:28<kwijibo>if you need a more generic class, there is dcmi:Dataset
20:32:12<fgiasson>maybe yes
20:32:22<kidehen>fgiasson: VoiD is not about LOD specifically, it is about HTTP accessible data entities (objects, items, resources). Of course the prime usecase today is the LOD cloud
20:34:19<kidehen>fgiasson: re. resource, it isn't a description of a "Thing" if its a darn non-information resource (a data object) . It is supposed to be a description of a "Thing" if doing Linked Data and you mean: information resource :-) I know you know this anyhow, so this is more about clarity based on my rewind of the thread
20:35:48<kidehen>fgiasson: kwijibo was trying to clarify the distinction and I noticed http-range-14 coming up etc..
20:36:53<kidehen>fgiasson: now for your main concern, I think VoiD should use sub-property approach for clarity instead of going directly to dct:isPartOf etc..
20:37:52<fgiasson>yeah; this is what I thought
20:39:34<kidehen>fgiasson: we have a collection of triples in a data set, but the collection is constellation of statements that describe a variety of subjects. There is a bona fide association between the subjects that a referenced in Subject and Object slots in a data set, and I think Toby came up with an elegant resolution for that.
20:40:27<kidehen>fgiasson: so as per usual everyone was quite right in this pretty long thread, but everyone (as per usual) was focused on their specific point of interest :-)
20:41:03<kidehen>fgiasson: my interest is typically peace and harmony :-)
20:41:13<fgiasson>:)
20:41:24<fgiasson>well; was just wondering; and maybe suggesting an enhancement; ehhehe
20:42:47<kidehen>fgiasson: worthwhile discussion, now just about harmonious progression of the suggestions etc.. :-)
20:42:56<kwijibo>i think, if there is to be a new property, the semantics i would want it to express would be something like: <foo.rdf> void:containsDataIn </meta#Dataset>
20:43:16<tobyink0>re isPartOf/hasPart - "A taxonomy of part-whole or meronymic relations is developed to explain the ordinary English-speaker's use of the term ?part of? and its cognates."
20:43:20<kwijibo>but preferably expressed as a noun
20:43:29<tobyink0>"The resulting classification yields six types of meronymic relations: 1. component-integral object (pedal-bike), 2. member-collection (ship-fleet), 3. portion-mass (slice-pie), 4. stuff-object (steel-car), 5. feature-activity (paying-shopping), and 6. place-area (Everglades-Florida)."
20:44:09<tobyink0>(that's from A taxonomy of part-whole relations by Morton E. Winston, Roger Chaffin and Douglas Herrmann.
21:03:37<kwijibo>tobyink: where is this from ?
21:03:44<kwijibo>tobyink0:
21:16:11<tobyink0>kwijibo - http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6W48-4F2T79V-N&_user=10&_rdoc=1&_fmt=&_orig=search&_sort=d&view=c&_acct=C000050221&_version=1&_urlVersion=0&_userid=10&md5=71d9e9cf8f5c24261addf0ffb0498f3d
21:32:22<Shepard>tobyink: I take it instance-class is not seen as a meronymic relation then?

Back to channel and daily index: content-negotiated html turtle