14:36:24 RRSAgent has joined #rdf-wg 14:36:24 logging to http://www.w3.org/2012/05/09-rdf-wg-irc 14:36:26 RRSAgent, make logs world 14:36:26 Zakim has joined #rdf-wg 14:36:28 Zakim, this will be 73394 14:36:28 ok, trackbot; I see SW_RDFWG()11:00AM scheduled to start in 24 minutes 14:36:29 Meeting: RDF Working Group Teleconference 14:36:29 Date: 09 May 2012 14:55:27 AZ has joined #rdf-wg 14:55:44 SW_RDFWG()11:00AM has now started 14:55:58 +??P3 14:55:58 + +31.20.598.aaaa 14:56:05 Guus has joined #rdf-wg 14:56:23 Zakim, who is on the phone? 14:56:23 On the phone I see ??P3, +31.20.598.aaaa 14:56:26 pchampin has joined #rdf-wg 14:56:27 Zakim, ??P3 is me 14:56:27 +yvesr; got it 14:57:07 zakim, +31.20 is me 14:57:07 +Guus; got it 14:57:19 zakim, dial ivan-voip 14:57:19 ok, ivan; the call is being made 14:57:20 +Ivan 14:57:38 zakim, this is RDF 14:57:38 Guus, this was already SW_RDFWG()11:00AM 14:57:39 ok, Guus; that matches SW_RDFWG()11:00AM 14:58:35 ScottB has joined #rdf-wg 14:58:47 +??P7 14:58:52 zakim, ??P7 is me 14:58:52 +AndyS; got it 14:59:14 PatH has joined #rdf-wg 15:00:04 +bhyland 15:00:13 Zakim, bhyland is really me 15:00:13 +davidwood; got it 15:00:28 +??P8 15:00:49 pfps has joined #rdf-wg 15:01:01 +OpenLink_Software 15:01:05 +Tony 15:01:09 zakim, ??p8 is me 15:01:09 +pfps; got it 15:01:13 ack ??p8 15:01:21 Zakim, Tony is temporarily me 15:01:21 +ScottB; got it 15:01:31 I'm the scribe, I'm joining the call 15:01:39 Zakim, OpenLink_Software is temporarily me 15:01:39 +MacTed; got it 15:01:40 Zakim, mute me 15:01:40 MacTed should now be muted 15:01:47 +??P15 15:01:52 +gavinc 15:01:56 tbaker has joined #rdf-wg 15:02:12 Zakim, ??P15 is me 15:02:12 +AZ; got it 15:02:30 I will be joining on IRC today but can call in if absolutely needed. 15:02:38 +??P19 15:02:53 +??P18 15:03:02 zakim, ??P18 is probably danbri 15:03:02 +danbri?; got it 15:03:03 zakim, ??P19 is tbaker 15:03:04 +tbaker; got it 15:03:30 scribe: AZ 15:03:35 +Sandro 15:04:10 +mhausenblas 15:04:14 zakim, mhausenblas is temporarily me 15:04:14 +cygri; got it 15:04:17 Topic: Admin 15:04:20 PROPOSED to accept the minutes of the 2 May telecon: 15:04:20 http://www.w3.org/2011/rdf-wg/meeting/2012-05-02 15:05:13 Review of actions 15:05:13 http://www.w3.org/2011/rdf-wg/track/actions/pendingreview 15:05:13 http://www.w3.org/2011/rdf-wg/track/actions/open 15:05:55 + +33.6.66.52.aabb 15:06:07 zakim, aabb is me 15:06:08 +pchampin; got it 15:06:30 Closed ACTION-168 as a duplicate 15:06:36 I've made no progress on any of mine, sorry. 15:07:41 Topic: XMLLiteral 15:07:47 q+ 15:07:55 See proposal at: 15:07:55 http://www.w3.org/2011/rdf-wg/wiki/XML_Literals 15:08:00 ack cygri 15:08:13 http://lists.w3.org/Archives/Public/public-rdf-wg/2012May/0006.html 15:08:17 +ericP 15:08:46 Zakim, who's noisy? 15:08:48 LeeF has joined #rdf-wg 15:08:56 MacTed, listening for 10 seconds I heard sound from the following: davidwood (8%), AZ (14%), ericP (15%) 15:09:14 scribenick: ericP 15:09:17 +LeeF 15:09:28 cygri: the lexical space need not be canonical, btu well-formed 15:09:43 • Make rdf:XMLLiteral optional in the datatype map 15:09:43 • Change rdf:XMLLiteral lexical space to allow 15:09:43 non-canonical but well-formed XML 15:09:43 • Define a canonical lexical form for rdf:XMLLiteral 15:09:43 that is equivalent to the old lexical space 15:09:44 • Re-define the value space in terms of XML infosets (this 15:09:46 should be in 1:1 correspondence to the old value space 15:09:48 AZ has joined #rdf-wg 15:09:48 and old lexical space) 15:09:49 ... then we can add a canonical lexical form, which is the same as the old lexical space 15:09:56 cygri: (describes rdf:XMLLiteral as found in link) 15:10:13 ... the value space would be 1:1 on the old values space, but we would want to rephrase the definition 15:10:21 ... there are two proposals: 15:10:31 ... .. expresses it in terms of infosets 15:10:41 http://lists.w3.org/Archives/Public/public-rdf-wg/2012May/0186.html 15:10:47 ... .. and we've just looked at expressing it in terms of DOM trees 15:10:57 ... DOM trees should be the same thing 15:11:28 q+ to ask why DOM (defined in terms of DOM) instead of infoset 15:11:33 -pchampin 15:11:39 q+ 15:11:52 cygri: question is how to define 15:12:19 ericP, you wanted to ask why DOM (defined in terms of DOM) instead of infoset 15:12:21 ack ericp 15:12:40 zwu2 has joined #rdf-wg 15:12:41 DOM is not phrased in term of the infoset 15:12:51 zakim, code? 15:12:51 the conference code is 73394 (tel:+1.617.761.6200 sip:zakim@voip.w3.org), zwu2 15:13:31 gavinc: XPath, XQuery, define their own data model 15:13:34 FYI: http://www.w3.org/TR/xml-infoset/#omitted 15:13:50 +zwu2 15:13:54 ack ivan 15:14:06 zakim, mute me 15:14:06 zwu2 should now be muted 15:14:33 gavinc: infoset has no conformance. all specs create their own model 15:14:39 A.isEqualNode(B) 15:14:42 ivan: we asked Liam, who said the same as gavinc 15:15:07 ... there is also a handy equiv function, A.isEqualNode(B), in DOM 15:15:30 ... another issue is whether we want to have an HTML5 literal 15:15:48 + +33.9.51.77.aacc 15:15:48 ... HTML5 is defines how to parse HTML5 into a DOM 15:15:53 zakim, aacc is me 15:15:53 +pchampin; got it 15:15:56 zakim, mute me 15:15:56 pchampin should now be muted 15:16:14 ... HTML5 does not go so far as how to say what HTML5 looks like in an infoset 15:16:31 ... we can chain specs to derive that, but it's complicated and unnecessary 15:16:35 q? 15:16:49 i'm happy for this choice as long as we have the blessing of Liam 15:17:01 davidwood: is this what steve harris objected to? 15:17:18 ivan: he had issues with the complexity 15:17:29 q+ 15:17:37 Ivan: We can define a path from an HTML5 literal to an infoset, but Steve had issues with that level of complexity in RDF. 15:17:38 ... but it's not required that one implement the equiv 15:18:02 FWIW, I am happy with anything as long as there is a well-defined literal-to-value mapping we can refer to in the semantics. 15:18:11 ... current defn demands that one create canonical XML 15:18:19 Ivan: Nobody knows what canonical XML is. 15:18:37 ... if you have a tool, like my RDFaDistiller, you're stuck finding a c14n library 15:18:46 s/Ivan:/Ivan,/ 15:18:48 q- 15:18:57 +q to add that it doesn't even have to be valid XML 15:19:17 ... so with the DOM soln, if tools want equality, they can use the DOM function 15:19:32 ack gavinc 15:19:32 gavinc, you wanted to add that it doesn't even have to be valid XML 15:20:04 gavinc: if defined in terms of DOM instead of XML C14N, we can leverage the HTML5 error handling 15:20:23 ... this can help us consume non-well-formed markup 15:20:36 s/HTML5/XML 15:20:43 q+ 15:20:57 See http://www.w3.org/community/xml-er/ 15:20:58 ack cygri 15:20:58 davidwood: richard's proposal exists in the context of needing of an XML datatype 15:21:08 ... so we can reduce the need for the XML datatype 15:21:26 s/needing of an XML datatype/needing of an HTML datatype/ 15:21:30 cygri: even if we don't change the effective datatype, a change to the defn makes it more usable 15:22:08 ... we're not ready to propose HTML literals, issues around parsing, etc 15:22:17 davidwood: but we've generally agreed that we'll do it 15:22:43 cygri: even before that, i propose redefining the XML literal 15:23:03 davidwood: make XML literal optional in the datatype map 15:23:12 +Arnaud 15:23:17 Change rdf:XMLLiteral lexical space to allow 15:23:17 non-canonical but well-formed XML 15:23:21 Arnaud has joined #rdf-wg 15:23:40 ivan: XML literals are not necessarily meant to capture HTML5 15:23:50 davidwood: we don't have an XHMTL type 15:24:04 XHTML is XML 15:24:12 HTML is HTML 15:24:21 ... hope 15:24:30 s/... hope / 15:24:55 Polyglut documents are funky and only crazy people like Sam Ruby make them 15:25:09 Wait. Good XHTML is XML< but can there be bad XHTML which is still good XML?? IF so, we need a separate datatype. 15:25:22 No, there is no such thing as "bad" XHTML 15:25:37 davidwood: regardless of what we do with XML and HTML datatypes, some data could go in either 15:25:56 Oh. Hmm, I guess I really should shut up at hthis point :-) 15:26:03 gavinc: "Polyglut" meaning a document that is both application/xhtml and text/html 15:26:07 ... those are hard to make 15:26:11 Polyglot too 15:26:31 s/Polyglut/Polyglot/ 15:26:52 I like polyglut. I knew one of them once. 15:27:34 q? 15:27:38 cygri: old XML value space is XML C14N, which specifies e.g. " vs. ', empty tags vs. tag pairs, etc. 15:27:39 q+ 15:27:44 ack ivan 15:28:06 ivan: do we need this canonical lexical form for each datatype? 15:28:08 Are there many users of rdf:XMLLIteral, in fact? 15:28:55 No, a dtatype does not *need* to hve a cononical form,. It just makes equality checking WAAAAY easier. 15:29:20 cononical/canonical 15:29:23 PatH, DOM defines equality checking 15:29:27 ericP: use cases for any canonicalization are around e.g. SPARQL queries looking for shoe:size 5 and not shoe:size "05"^^xsd::integer 15:29:30 Users - yes and no. GML literals are XML (but often not legal XMLLiterals) 15:29:31 Well then fine. 15:29:44 ... use cases for the XML Literal analog are a little bit of a stretch 15:29:50 q+ 15:30:18 ivan: responding to PatH, DOM-level equiv is easier than C14N equiv 15:30:48 OK. 15:30:50 ack AndyS 15:31:13 q+ to ask how equiv is used in anger 15:31:49 AndyS: it's clear how canonicalization is used 15:32:11 If you speak to me like that again., I'll equiv you so fast you won't know you've been canonicalized. 15:32:20 ... c14n is more in favor of containing the complexity to input normalization 15:32:47 ... unfortunetely, many XML literals can't just be pasted 15:33:02 ... you've moved the problem to someone else 15:33:23 +1 ericP 15:33:37 q+ 15:33:37 Or was it Andy. 15:33:43 davidwood: ok to push to someone else if the string is to be interpreted in someone else's application 15:33:56 Gotcha 15:34:27 gavinc: isn't there a clear optomization path? 15:34:42 q+ 15:34:52 ack ericp 15:34:52 ericP, you wanted to ask how equiv is used in anger 15:35:11 AndyS: depends on whether you want the output to exactly reflect the input 15:35:46 ... i'd like to encourage folks to canonicalize, but not oblige them 15:36:18 davidwood, if we do input normalization, they incur a cost for a data element which may never be read [or matched -- ED] 15:36:24 davidwood: if we do input normalization, they incur a cost for a data element which may never be read [or matched -- ED] 15:36:51 Seems to me key issue is, if I don't coninicalize, will your queries work right against my data? And if not, whose fault is that? 15:36:58 ... if you canonicalize on use, e.g. SPARQL, we impact those apps 15:37:25 zakim, unmute me 15:37:25 pchampin should no longer be muted 15:37:26 ack pchampin 15:37:28 pat, if the query engine implements equality of the dom trees, then it should work 15:37:39 ... it seems easier technically and socially to canonicalize on input 15:37:58 pchampin: i agree with AndyS's point 15:38:18 ... prob is folks won't necessarily know it's canonicalized and not take advantage 15:38:22 Ivan, OK, then why are we discussing canonicalizing on input? 15:38:27 ack cygri 15:38:33 ... an option is to have two, one with a restricted lexical space 15:38:34 cygri 15:38:39 pat, I do not know, that was i was asking, too! 15:38:44 cygri: that's what i proposed, but it's not working out 15:38:50 Ah, pchampin makes good point. 15:39:00 +1 to Richard 15:39:11 ... i don't think that requiring canonicalization has worked out 15:39:18 c14n in xml literals has proven to be a disaster 15:39:29 q+ 15:39:33 pchampin: may it didn't work out 'cause it was the only one available 15:39:35 Having a normative requirement to play fair rarely works out. 15:39:35 I agree requiring canonicalization has not worked out. 15:39:48 +1 to pchampin's point 15:39:54 zakim, mute me 15:39:54 pchampin should now be muted 15:40:17 -AZ 15:40:21 q+ 15:40:35 ivan: technically, two types could work, but i don't see the motivating use cases 15:40:52 ack ivan 15:40:57 ... mostly i've seen e.g. excerpts of HTML in RSS, used only for display 15:41:09 ... i don't think we should define another form of these datatypes 15:41:27 @ivan: fair enough :) 15:41:27 ... anyone could add that type 15:41:31 q? 15:41:36 ack cygri 15:41:40 cygri: +1 to ivan 15:41:59 ... it might be useful to keep the definition of the canonical mapping in the datatype 15:42:22 ... XS datatypes optionally define canonical forms 15:42:55 ... it's nice to indicate how c14n can be used by interested systems 15:43:11 If caonicalized data smells the same as uncanonicalized, then nobody can rely on the canon, so its not worth doing the work to canonicalize it, so we have a huge negative feedback situation. 15:43:12 ... there's no obligation, and for some systems it's useful 15:43:26 ... i think that pointing to the c14n algorithm is a good idea 15:44:22 ... also helps migration of RDF2004 to RDF1.1 by saying that the new lexical space encompasses the old space 15:44:55 Maybe have a datatype for canonicalized data? rdf:CXMLLIteral, to let people know what they are getting. 15:46:10 ericP, no, it doesn't mention unicode normalization 15:46:13 PROPOSED: in RDF 1.1 (a) XMLLiterals are optional (b) lexical space consists of valid XML fragments © the canonical lexical form is c14n (d) the value space consists of DOM trees 15:46:35 s/valid/well-formed/ 15:46:48 Q: is this harder than the current situation or easier? 15:47:13 And for who? (publishers or consumers?) 15:47:17 Yes. 15:47:19 easier is good! :-) 15:47:25 I'm with pfps. 15:47:28 Are DOM trees unique? 15:47:42 Easier to publish and easier to consume 15:47:45 zwu2: not necessarily 15:47:54 I like that this doe not use the word "canonicalize" 15:47:57 only after normalization 15:47:59 then, which one should we canonicalize into? 15:48:13 +1 15:48:17 1+ 15:48:19 Looks best of (difficult) choices to me. 15:48:19 q+ 15:48:27 ivan: i am much more interested in keeping publishing easier 15:48:33 ack ericp 15:48:44 +1 ivan 15:48:45 +1 ivan: we should make it easier for data publishers, even if it makes things harder for SPARQL implementers 15:48:52 (that is, +1 make publishing easier, even at the expense of making SPARQL/consumption harder) 15:49:03 q+ 15:49:04 q+ to address the SPARQL aspect 15:49:30 +q to point out that SPARQL stores can still use C14N 15:50:38 ack ivan 15:51:19 ivan: in RDFa, the test harness uses SPARQL ASK to test a particular pattern 15:51:41 -q 15:51:44 ... we had immense problems with the SPARQL literals, uneven implementation 15:51:57 ack AndyS 15:51:58 AndyS, you wanted to address the SPARQL aspect 15:51:59 ... it's aleady a mess; we won't make it worse 15:52:21 s/SPARQL literals/XML literals in SPARQL/ 15:52:22 AndyS: i think the SPARQL stores would handle it at load time instead of query time 15:52:42 would entailment permit that? 15:53:27 You end up building hashes based on the DOM itself and the XPath/XQuery data model, see http://exist-db.org/ ;) 15:53:29 q? 15:53:33 [ discussion of errors in large uploads ] 15:53:52 q? 15:54:18 2012-05-09T15:46:13Z 15:54:20 PROPOSED: in RDF 1.1 (a) XMLLiterals are optional (b) lexical space consists of valid XML fragments © the canonical lexical form is c14n (d) the value space consists of DOM trees 15:54:31 q+ 15:54:46 zakim, unmute me 15:54:46 zwu2 should no longer be muted 15:54:49 What does © mean here? 15:54:56 davidwood: i think no one objects to XMLLiteral are optional or lexical space consists of valid XML 15:54:57 PatH - client error 15:55:01 +1 to a, b, c, and +∞ to d 15:55:05 \me (c) 15:55:11 pat: my client turned ( c ) into a copyright character:-( 15:55:18 PROPOSED: in RDF 1.1 (a) XMLLiterals are optional (b) lexical space consists of valid XML fragments (c) the canonical lexical form is c14n (d) the value space consists of DOM trees 15:55:20 AH. Duh. 15:55:28 PROPOSED: in RDF 1.1: [a] XMLLiterals are optional; [b] lexical space consists of valid XML fragments; [c] the canonical lexical form is c14n; [d] the value space consists of DOM trees. 15:55:30 \me that's because it *knew* that you worked for W3C 15:55:41 *heh* 15:55:44 q+ 15:55:59 ack zwu 15:56:34 I wonder what is the point of stating that there is a canonical lexical form if peple arent obliged to use it and users can't tell if it has been used or not. 15:56:41 zwu, i like the idea of c14n, but is c14n + serialization a unique process? 15:56:45 Zakim, who's noisy? 15:56:55 q? 15:56:56 MacTed, listening for 10 seconds I heard sound from the following: Guus (15%), davidwood (29%), Ivan (55%), zwu2 (15%), ericP (35%) 15:57:10 I want to know if c14n + serialization a unique process 15:57:27 actually ericP captured my question, thanks! 15:57:46 A.isEqualNode(B) dom3 15:57:56 ivan: we don't care, 'cause what counts is the equality in the value space 15:58:32 Arnaud: that equivalence is post-normalization 15:58:55 ... e.g. creating a single text node from a series of text nodes 15:58:56 q? 15:58:57 -pchampin 15:59:01 I take it that the only purpose of mentioning a canonical form is so that equality reduces to string identity (or close) . If this is not an issue, then lets just not even mention canonicalization. 15:59:10 cygri: not needed 'cause the DOM tree is the result of parsing 15:59:13 PatH, yes 15:59:30 +1 to PatH 15:59:37 Pat that was my point... 15:59:37 -pfps 15:59:51 Ivan, then delete ( c ) 16:00:09 Arnaud: that's not defined 16:00:24 ericP: i've seen the opposite from MSXML3 16:00:44 +??P6 16:00:49 q+ 16:00:52 zakim, ??p6 is me 16:00:52 +pfps; got it 16:00:55 davidwood: we could have this same discussion based on, say, a style modification 16:00:58 ack ivan 16:01:10 ... we can never solve this, just assympotically approach it 16:01:27 Davidwood, Oooh yes, lets! 16:01:28 We can lean on DOM anyway here, or reuse the wording ;) "The childNodes NodeLists are equal. This is: they are both null, or they have the same length and contain equal nodes at the same index. Note that normalization can affect equality; to avoid this, nodes should be normalized before being compared." 16:01:34 ivan: HTML5 spec is very clear about how a document is turned into a DOM 16:01:55 zakim, mute me 16:02:02 +pchampin 16:02:13 pchampin should now be muted 16:02:22 ericP: does HTML5 produce exactly one DOM? 16:02:22 see http://www.w3.org/TR/DOM-Level-3-Core/core.html#ID-normalize 16:02:28 ivan: assume so 16:02:59 ... content to take Arnaud's advice about normalizing first 16:03:37 PROPOSED: in RDF 1.1: [a] XMLLiterals are optional; [b] lexical space consists of well-formed XML fragments; [c] the canonical lexical form is c14n; [d] the value space consists of DOM trees. 16:04:09 [d] the value space consists of (normalized) DOM trees. 16:04:34 PROPOSED: in RDF 1.1: [a] XMLLiterals are optional; [b] lexical space consists of well-formed XML fragments; [c] the canonical lexical form is c14n; [d] the value space consists of (normalized) DOM trees. 16:04:38 TO me, "canonical" is easy to read as meaning "recommended". Do we want to convey this? 16:04:56 abc could be element(root, (text node(a),text node(b),text node(c)) or element(root, (text node(abc)) 16:05:10 (XSD defines canonical forms -- does not force use) 16:05:24 +1 16:05:27 +1 16:05:29 +1 16:05:30 +1 16:05:30 +1 16:05:32 +epsilon 16:05:33 +1 16:05:33 +1 16:05:33 -1 to [c] 16:05:34 -1 16:05:37 +1 16:05:42 +1 16:05:44 +0 16:05:55 that was -1 to [c], +1 to rest. 16:06:22 Doesn't the current situation require canonicalization? 16:06:26 Yes. 16:06:37 PROPOSED: in RDF 1.1: [a] XMLLiterals are optional; [b] lexical space consists of well-formed XML fragments; [c] the value space consists of (normalized) DOM trees. 16:07:32 ericP, why don't you ask them 16:07:35 I am fine with canonicalization which is REQUIRED. BUt if its not required, we shouldnt mention it at all. 16:07:43 -0.5 to not mentioning what the canonical form is : we are suggesting canonical for integers etc as good practice. 16:08:00 Good practive is fine, but not in the definitions. 16:08:42 Jena checks - can't remember is it will canonicalize - maybe does it by string->DOM->string 16:08:43 PatH, ALL the xsd datatypes define a cononical form 16:08:44 Does producing (normalized) DOM trees require canonicalization? 16:08:48 http://www.w3.org/TR/xmlschema-2/#boolean 16:08:53 pfps, no 16:09:22 Gavin,. I know it is defined. The issue is, do we require its use in RDF data? If not, lets not mention it in the normative definition of the datatype. 16:09:27 Non-normative section. 16:09:53 non-normative refrence to cononical form 16:10:00 Exactly 16:10:00