Nov 02 11:44:11 sbp, are you interested to review the GRDDL mime type/namespace situation? Nov 02 11:44:48 certainly! Nov 02 11:44:54 perhaps the best place to start is http://lists.w3.org/Archives/Public/www-tag/2006Mar/0008 Nov 02 11:45:02 oops; rather http://lists.w3.org/Archives/Public/public-grddl-wg/2006Nov/0006.html Nov 02 11:46:08 one way to phrase the question is: does http://www.w3.org/2001/sw/grddl-wg/td/sq1ns assert any RDF triples? or is it just a bunch of XML elements and attributes? Nov 02 11:46:31 it has only one representation, labelled application/xml Nov 02 11:46:56 interesting question. I was just looking through RDF Syntax to see what it says... Nov 02 11:47:04 --> `miguel (n=chatzill@64.157.92.85) has joined #swig Nov 02 11:47:13 'The Internet media type / MIME type for RDF/XML is "application/rdf+xml"' - Section 4, http://www.w3.org/TR/rdf-syntax-grammar/ Nov 02 11:47:35 that doesn't appear to be a conformance constraint though Nov 02 11:48:04 yes, re what the RDF spec says, what the OWL spec says is also relevant... http://lists.w3.org/Archives/Public/public-grddl-wg/2006Nov/0011.html Nov 02 11:48:17 oh, "An RDF/XML Document is a conforming RDF/XML document if it adheres to the specification defined in this document." Nov 02 11:49:05 ah, that's good. of course there's nothing to prevent the W3C from coming up with a new media type and making it a subset of application/rdf+xml... as could have happened with OWL Nov 02 11:50:22 but then the whole point of XML is that it lets you dispatch on syntax -> meaning hooks with some granularity Nov 02 11:50:35 if the MIME type is a generic XML one, then you look at the namespace of the root element Nov 02 11:50:57 "if the MIME type is a generic XML one, then you look at the namespace of the root element" <- that's a coherent position, but it's not one I can justify from ratified specs Nov 02 11:50:58 then you look at the specification for that namespace, and it might delegate to other specifications if it allows foreign content Nov 02 11:51:17 ah, I thought that the TAG were working on that as part of the whole XML Functions thing Nov 02 11:51:23 but I admit that I'm not up to date with that Nov 02 11:51:24 "if the MIME type is a generic XML one, then you look at the namespace of the root element" <- that's a "yes" answer in the GRDDL WG poll of 1 Nov. Nov 02 11:51:29 * DanC needs to get those minutes out Nov 02 11:51:40 yes, exactly, XML functions is open Nov 02 11:51:43 * chimezie catches up on juicy scrollback Nov 02 11:52:03 --> leobard (n=Miranda@pc-215.dfki.uni-kl.de) has joined #swig Nov 02 11:52:22 ah, and mixedNamespaceMeaning-13 Nov 02 11:52:42 the RDF spec doesn't help here. the definition of RDF/XML document is in terms of RDF Document, and the whole question is whether http://www.w3.org/2001/sw/grddl-wg/td/sq1ns is an RDF document. Nov 02 11:52:56 * sbp sees tagIssuesTaxonomy-55 on the horizon... Nov 02 11:53:07 the TAG decided issue 13 is equivalent to issue 42 and sorta withdrew it. Nov 02 11:53:17 heh Nov 02 11:53:44 oh, I see the subsumedness now Nov 02 11:54:28 i think the issue has to do more with web architecture and XML (as pure syntax) than what RDF might mandate for a mime-type Nov 02 11:54:58 --> DanC_lap (n=connolly@64-126-89-30.dyn.everestkc.net) has joined #swig Nov 02 11:55:11 well not really, since RDF parsers are likely to conform to the RDF Syntax specification Nov 02 11:55:22 and if it says to reject non-application/rdf+xml, then they'll reject it Nov 02 11:55:29 yeah, but this is a question of what you are parsing before you even know what it is Nov 02 11:55:57 yeah, there should be enough information boasted in the headers to find out Nov 02 11:56:07 similar sort of question applies to XHTML and the profiling going on there Nov 02 11:56:13 but if there is none .. Nov 02 11:56:26 it could be important for, say, small devices to know whether a mobile profile is being used, say XHTML Basic Nov 02 11:56:31 sbp, chimezie does have a point about this being more than an RDF issue; suppose you get a representation labelled application/xml with the SVG namespace on the root element. Should you draw circles? Nov 02 11:57:47 I don't think the RDF spec says to reject non-application/rdf+xml. I think it's just silent on documents labelled application/xml Nov 02 11:57:48 again, I'd say it depends on what the SVG specification says; modulo the fact that you used *should*. I think I'm taking the more descriptive line: trying to model how software programmers work; they read the specs that they need to read and implement to them Nov 02 11:58:19 and if there are conformance rules mandating the use of a single particular media type for some format, then the code that they write is going to reject all other forms as being non-conformant Nov 02 11:58:43 how do we get as far as the SVG specification? as to should, let me rephrase: may I draw circles and attribute them to the document's author? Nov 02 11:59:07 E_UNKNOWN Nov 02 11:59:14 hehe Nov 02 11:59:44 I can imagine going to court over it Nov 02 12:00:01 court is in session. the jury is called the GRDDL WG. Nov 02 12:00:02 and pleading in front of the judge that I was using application/xml because I wanted the semantics to be opaque but the infoset to be accessible Nov 02 12:00:27 * chimezie remembers the question he posed about about the legal authority of the different ways you might embed creative commons statements in a page Nov 02 12:01:00 suppose you're on the jury, sbp. which way do you vote? Nov 02 12:01:04 well on the other hand I could argue that the granular dispatch chain is well known. I'll bet I could find code that goes a-namespace sniffin' Nov 02 12:01:06 --> Kriegel (n=chatzill@141.76.46.190) has joined #swig Nov 02 12:01:21 indeed, cwm and rdflib do go a namespace-sniffin Nov 02 12:01:43 I vote that at the moment you have to do what all good jurists would do in any case: look for intent on behalf of the author via other methods. if you can't find any evidence, then you must acquit Nov 02 12:01:58 namespace sniffing seems to suggest that namespace uri's should be intepreted as identifiers *first* before being treated as locations Nov 02 12:02:03 but were I judge, I'd call for laws to be made on the subject Nov 02 12:02:15 because that seems to be the problem: we're saying, what is the law here? and there isn't any Nov 02 12:02:25 on cwm and rdflib sniffin: http://lists.w3.org/Archives/Public/public-grddl-wg/2006Nov/0007.html Nov 02 12:02:30 is there a commonly used RDF application to describe projects in a company? I'm hoping for something that tie into FOAF. Nov 02 12:03:09 hmm... sbp, I just can't get you to give an opinion, can I? the judge can't wait for the legislature; his job is to see that justice is done today Nov 02 12:03:18 ah, interesting. and I've written code that accepted *text/plain* as RDF, using some major-mojo content sniffing before, because I found that lots of people were using text/plain for Notation3 and even RDF/XML. it was horrible Nov 02 12:03:36 yeah, which is why in the particular case before you you have to rule on intent via whatever evidence possible Nov 02 12:03:55 the analogy breaks down here because this isn't a transgression Nov 02 12:04:02 so you can't default to innocent Nov 02 12:04:03 hmm... actually, it's probably a better analogy to say that the GRDDL WG is the legislature. Nov 02 12:04:57 which law makes for a better society? the one that says "if you use application/xml and the RDF namespace, you're accountable for the RDF statements therein"? or the one that says "if you want to hold somebody accountable for RDF statements, you have to find a representation labelled application/rdf+xml"? Nov 02 12:04:59 .gc grddl Nov 02 12:05:01 * sbp remembers when CWM used to try to identify RDF/XML by sniffing for "xmlns" in the content... until danbri registered xmlns.com for FOAF! Nov 02 12:05:02 grddl: 1,370,000 Nov 02 12:05:08 crikey! Nov 02 12:05:25 * DanC wonders if there's some other meaning of the word "grddl" Nov 02 12:05:35 doesn't appear so! that is indeed surprising Nov 02 12:05:49 .ety grddl Nov 02 12:05:51 Can't find the etymology for "grddl". Try http://etymonline.com/?search=grddl Nov 02 12:05:52 lol: re: xmlns.com breaking sniffing ;) Nov 02 12:05:58 if I were on the legislature, I guess I'd want to do a survey of what people are assuming at the moment Nov 02 12:06:15 my prediction would be that it would end up being the former of your two options Nov 02 12:06:27 perhaps by a huge majority Nov 02 12:06:28 aha! you emitted the bit! Nov 02 12:06:39 problem with sniffing namespace uris as locations is that alot of the major vocabulary namespace uri's dont resolve to anything useful, but have an authoratitive identifier Nov 02 12:06:41 1 Nov 02 12:07:29 chimezie: well, a lot of the major media types don't resolve to anything useful... :-) Nov 02 12:07:42 * DanC has trouble understanding the relevance of the location/identifier distinction here Nov 02 12:07:59 yeah, I think it's irrelevant Nov 02 12:08:27 I recall sbp you had a suggestion before that transforms (per GRDDL) should be interpreted as identifiers Nov 02 12:09:01 i.e., if the author refers to a common transform, the processor shoudl not have to dereference to get the transform to use, it should match the identifier against a 'registry' Nov 02 12:09:09 --> kandinski (i=kandinsk@rowrcolo.net) has joined #swig Nov 02 12:09:11 yes, absolutely Nov 02 12:09:18 if you consider the rdf namespace as a unique identifier for XML content.. then it's a similar scenario Nov 02 12:09:24 yes, quite, chimezie . but how does that relate to the choice of which law makes a better society? Nov 02 12:09:39 and you can link that back to all sorts of stuff; for example, I remember Dave Pawson expressing disbelief that not more people had SGML catalogs Nov 02 12:09:40 no need to *go* to the rdf namespace (i.e., resolve as a location) , just parse it as RDF Nov 02 12:10:02 in neither case do we dereference the RDF namespace URI Nov 02 12:10:13 what is at the rdf namespace uri currently? Nov 02 12:10:30 * chimezie tries Nov 02 12:10:32 a hunk of RDF that sorta declares each of the terms in the namespace Nov 02 12:10:38 yeah, Content-Type: application/rdf+xml Nov 02 12:10:59 but again, whilst it's an interesting issue... how does it impact upon the choice here? Nov 02 12:11:17 --> mlinksva (n=mlinksva@pdpc/supporter/sustaining/mlinksva) has joined #swig Nov 02 12:11:57 well, consider if a 'processor' was mandated to attempt to parse the document as RDF if the namespace uri was the rdf namespace then you would essentially be telling it to respond as though the namespace uri was an identifier Nov 02 12:12:28 right. but since the choice is between namespace URI as an identifier and media type as an identifier... Nov 02 12:12:29 rather than - as it is now - where it would follow it's nose (unecessarily) Nov 02 12:12:33 it doesn't bear upon the question Nov 02 12:12:52 well, the assumption here is that you *don't* have a media type Nov 02 12:13:01 nope. the question is whether we should make only application/rdf+xml documents be those which are RDF/XML documents, or include appliation/xml + RDF namespace documents *too* Nov 02 12:13:13 sure you have a media type: application/xml Nov 02 12:13:35 right, but it isn't authoritative about how you handle the content Nov 02 12:13:35 this wouldn't apply if it were text/plain, for example Nov 02 12:13:41 right Nov 02 12:13:54 so the question here is: should it provide some algorithm about that? Nov 02 12:13:58 or should it not? Nov 02 12:14:03 <-- DanC_lap has quit (Read error: 110 (Connection timed out)) Nov 02 12:14:31 and if it doesn't, does that mean that the document is to be interpreted as just a kind of reified infoset with no meaning beyond being an interesting datastructure? Nov 02 12:14:57 well, I'm suggesting that the currently sanctioned algorithm (or suggestion) is to go to the namespace document (as a location) Nov 02 12:15:09 okay. sanctioned where? Nov 02 12:15:31 oh, you mean that you think that to be the best current practice? I see Nov 02 12:15:42 it's the best practice the TAG is pushing Nov 02 12:15:52 and the motivation for the namespace sniffing that GRDDL currently mandates Nov 02 12:16:04 same goes for RDDL Nov 02 12:16:34 ah right, yes... but... Nov 02 12:16:37 it's like another step Nov 02 12:17:06 we've got the Media Type vs. Namespace step, and then if we settle on Namespace we've got the Identifier vs. Location vs. Both/Either/WhoCares step Nov 02 12:17:12 and we have to answer the former before the latter, clearly Nov 02 12:17:30 yes, definately in that order Nov 02 12:17:40 which is why it's irrelevant *now*, even though interesting and will be the next thing to consider, definitely Nov 02 12:17:55 you're skipping ahead, dude! :-) Nov 02 12:18:00 :) Nov 02 12:18:13 <-- leobard has quit (Connection timed out) Nov 02 12:18:48 I'd still like to do a real survey of how tools treat application/xml Nov 02 12:19:39 would you want to look at SVG tools too? or just RDF/OWL tools? Nov 02 12:20:16 any tool which groks application/xml, since I think you'd have to make the decision at the application/xml specification level Nov 02 12:20:23 because in some sense, when you fire up an RDF tool, you the consumer are saying "never mind what the author told the web; I'm willing to take the risk that he meant RDF" Nov 02 12:20:47 true. but then, that's kinda getting into screen-scraping almost Nov 02 12:20:52 exactly Nov 02 12:20:55 which is anarchitectural Nov 02 12:21:03 heh... that's a damn good prefix pun Nov 02 12:21:37 the whole question here is whether looking at namespace names in application/xml is scraping or good architecture. Nov 02 12:22:48 I think it's good architecture provided the GRDDL spec goes thru due process. i.e. we _are_ taking away some liberties, and we need the consent of the goverened to do that. Nov 02 12:23:10 I'd say it's undetermined now, but should be considered good architecture, and normatively documented as such, due to the amount of current support (i.e. the number of people just assuming it or doing it anyway) Nov 02 12:23:32 "people do it anyway" isn't exactly a watertight argument, but probably better than no argument, which is what we had before considering this :-) Nov 02 12:23:51 a tough question because what you describe spans the state of the web, the headers you get, the difference between RDF/XML as syntax and as 'semantics', etc.. Nov 02 12:24:12 GRDDL spec: so you're saying that the normative documentation of it being a good architectural practice could be in the GRDDL specification? I fail to see how Nov 02 12:24:28 for me it should naturally belong in the application/xml media type specification, or whatever it defers to Nov 02 12:24:41 which would be much, much more tricky to change of course... Nov 02 12:24:58 (OT and Advert! http://food.yahoo.com/ just launched, the editorial content management is all in rdf & redland, rdf/owl schema, metadata queried with sparql, edited with xforms, ...) Nov 02 12:25:01 but where does one look for the meaning of an application/xml document if not the application/xml media type specification? Nov 02 12:25:16 --> shellac (n=pldms@82-32-5-17.cable.ubr01.azte.blueyonder.co.uk) has joined #swig Nov 02 12:25:17 edited in xforms.. Nov 02 12:25:28 nice Nov 02 12:25:30 doesn't sound OT to me Nov 02 12:25:46 if any specification can bear on any technology, one would have to read every single specification before implementing anything. (probably not a bad idea in some ways, but...) Nov 02 12:25:58 didn't want to derail your conversation, but the site went live today, carry on. Nov 02 12:25:59 which actually, is kinda already true for the XML Core suite Nov 02 12:25:59 * DanC hopes to find some way around sbp's argument that the applicatoin/xml spec needs to be changed Nov 02 12:26:10 * sbp hopes you can too... Nov 02 12:26:19 * sbp looks at food.yahoo Nov 02 12:26:19 about time xforms & rdf took off Nov 02 12:26:54 awesome. now all you need is a "Powered by Redland" link at the bottom! :-) Nov 02 12:27:20 <-- LotR_ has quit (Read error: 60 (Operation timed out)) Nov 02 12:27:25 i would consider it good architecture on the strenght of the precedent that http://www.w3.org/2001/tag/doc/nsDocuments/ sets Nov 02 12:27:36 s/strenght/strength Nov 02 12:27:37 is the SPARQL endpoint for food.yahoo public, by chance? Nov 02 12:28:30 there's no endpoint - it's just query at present, inside Nov 02 12:28:46 hmm... "This finding addresses TAG issue namespaceDocument-8." issue 8 is not yet resolved. Nov 02 12:28:58 yeah. Draft Finding Nov 02 12:29:00 really? Nov 02 12:29:06 * chimezie was under the impression it was Nov 02 12:29:12 "Draft TAG Finding 13 December 2005" Nov 02 12:29:37 http://www.w3.org/2001/tag/issues says no decision yet Nov 02 12:29:56 sorry, chimezie , the TAG should be more clear about such things. Nov 02 12:30:05 and good thing too; Norm has Things To Do :-) Nov 02 12:30:36 * sbp went through tag/issues in a fair amount of detail just a day or two ago and stumbled over the clarity and lack of links sometimes Nov 02 12:30:42 meanwhile, http://www.w3.org/TR/webarch/#pr-namespace-documents is ratified... Nov 02 12:31:01 I think that reflects the status of the issues nonetheless though; some seem to have just dropped off the radar for no particular reason Nov 02 12:31:08 "Nadia receives representation data from "weather.example.com" in an unfamiliar data format. She knows enough about XML to recognize which XML namespace the elements belong to." Nov 02 12:31:36 it's still on the radar; norm's action has been continued explicitly in recent weeks Nov 02 12:31:42 * sbp reads Nov 02 12:33:12 --> juvi (i=jukka@jt11-141-2.tky.hut.fi) has joined #swig Nov 02 12:33:35 (incidental note: siteData-36 isn't mentioned as being related in http://lists.w3.org/Archives/Public/www-tag/2003Jul/0127 [for rdfURIMeaning-39], which seems an oversight. another thing close on my radar) Nov 02 12:34:47 * sbp wonders what status "Story" has in a TAG authored W3C recommendation... Nov 02 12:34:51 I don't see any particular connection between rdfURIMeaning-39 and siteData-36 Nov 02 12:34:54 <-- anselm has quit ("Terminated with extreme prejudice - dircproxy 1.0.5") Nov 02 12:35:20 DanC: consider it an action item on me to point out my perceived connection on www-tag Nov 02 12:37:48 hmm wait, there is precedent for this sort of thing. the relative namespaces decision at the XML Plenary Nov 02 12:38:04 I was just thinking whether extra-XML Core documents can mandate on XML issues Nov 02 12:39:28 on the other hand, it seems like the way that that was implemented was by trying to influence future specifications to make relative namespaces meaningless Nov 02 12:39:51 * DanC pores over http://www.rfc-editor.org/rfc/rfc3023.txt <- http://www.iana.org/assignments/media-types/application/ Nov 02 12:40:27 "xmlns" doesn't occur :-/ Nov 02 12:40:46 ooh! ooh! [[ Nov 02 12:40:47 An XML document labeled as text/xml or application/xml might contain Nov 02 12:40:47 namespace declarations, stylesheet-linking processing instructions Nov 02 12:40:47 (PIs), schema information, or other declarations that might be used Nov 02 12:40:47 to suggest how the document is to be processed. Nov 02 12:40:47 ]] Nov 02 12:40:58 I guess I cited that in the OWL discussion. Nov 02 12:41:16 it comes down to interpretation of "might" and "suggest" Nov 02 12:41:17 ooh indeed! Nov 02 12:41:48 nice!! Nov 02 12:41:59 now that's authoritative Nov 02 12:42:02 "Such a document might be handled by applications that Nov 02 12:42:02 would use this information to dispatch the document for appropriate Nov 02 12:42:02 processing." Nov 02 12:42:19 _appropriate_ seems to endorse following ones nose Nov 02 12:44:40 hmm. but isn't our question whether some application/xml document asserts some triples? it seems that RFC 3023 just says "user agents may take the document to be asserting triples", but that doesn't seem to bear on *concretely* understanding the author's intent Nov 02 12:44:42 on the other hand! Nov 02 12:44:43 --> anselm (n=anselm@64.71.151.167) has joined #swig Nov 02 12:44:52 if you're publishing RDF/XML as application/xml... Nov 02 12:44:59 and you have read RDF 3023... Nov 02 12:45:16 then you ought understand that some user agents may interpret you as having asserted the triples in the RDF/XML Nov 02 12:45:23 therefore, you are asserting them, it seems to me Nov 02 12:45:37 well, I wonder how much room there is for misinterpeting the intent of a document that was served as application/xml but has an rdf namespace uri at the root Nov 02 12:45:49 how likely is it that the author *didn't* intend for it to be parsed as rdf? Nov 02 12:46:19 chimezie: from this reading of RFC 3032, it appears not at all Nov 02 12:46:33 that if they didn't intend for it to be parsed as RDF, they shouldn't've published it in such a manner Nov 02 12:46:41 yes Nov 02 12:46:47 and that if they intended otherwise, then they didn't understand application/xml Nov 02 12:47:00 either way they're liable Nov 02 12:47:12 I read it as saying that it's appropriate for a user agent to attribute the triples to the author. Nov 02 12:47:37 indeed, "therefore, you are asserting them" Nov 02 12:47:58 loggy, pointer Nov 02 12:48:02 you agree that to be a corollary? excellent Nov 02 12:48:20 I'll get that someday Nov 02 12:48:32 <-- terraces has quit ("Lost terminal") Nov 02 12:48:33 heheh. wrong channel Nov 02 12:48:48 huh? no logger? Nov 02 12:48:50 and oh dear, all that was unlogged? Nov 02 12:48:57 oh boy! Nov 02 12:49:04 * chimezie glares at dajobe Nov 02 12:49:08 hey! Nov 02 12:49:26 no fair griping when a free service suffers an occasional fault Nov 02 12:49:34 yeah, and we were just as likely to notice as him Nov 02 12:49:49 oh I was just pulling his chain Nov 02 12:49:51 at any rate, I suggest that one of us emails logs to www-archive Nov 02 12:49:52 I have a local log Nov 02 12:49:52 * kandinski brings out pen and paper Nov 02 12:50:08 I nominate DanC since he was first to admit to having local archives :-)