W3C RDF Core Working Group IRC Chat Logs for 2001-10-05

13:01:03 Topic now RDF Core WG - Syntax Meeting 14:00 UTC Friday (Usual Time)

13:01:03 Users on #rdfcore: logger_1 @dajobe

13:07:59 <dajobe> logger_1, off

13:26:46 <dajobe_1> dajobe_1 is now known as dajobe-batcave

13:27:14 <dajobe> logger_1, off

13:48:16 <dajobe-batcave> dajobe-batcave is now known as dajobe-jan

13:50:58 <DanC> hmm... no ops

13:59:56 <dajobe> dialled in...

14:02:21 * DanC is running just a bit late...

14:02:36 <ArtB> What's the number?

14:02:56 <dajobe> +1 630 536 3003 room #3003

14:03:56 <ArtB> +ArtB

14:04:53 <dajobe> Agenda

14:05:07 <dajobe> 0 scribe

14:05:09 <dajobe> 1 aob

14:05:17 <dajobe> 2 schema validation technology

14:05:25 <dajobe> 3 transformational aprproaches

14:05:37 <dajobe> 4 outstanding issues from issues list

14:05:47 <dajobe> --

14:06:29 <dajobe> jang to scribe

14:06:36 <dajobe> dajobe is now known as scribe-jang

14:06:45 <scribe-jang> 2 expanded:

14:06:51 <scribe-jang> what tech do we use to validate?

14:06:54 <scribe-jang> 3 expanded

14:06:55 * DanC would like a brief statement of a/the goal for this meeting

14:07:14 <scribe-jang> what transformational approaches tdo we take? english? functional vs declarative...

14:07:27 <scribe-jang> 4. any issues still on the issues list that are schema issues?

14:07:51 <scribe-jang> what about work that's been done? (danc)

14:08:22 <scribe-jang> item 2

14:08:28 <scribe-jang> relaxng seems popular

14:08:38 <DanC> ------------------ schema validation technology

14:08:40 <scribe-jang> nods all round from daveb, jang, jeremy

14:08:54 <scribe-jang> do we agree a normative validation tech is a good idea?

14:09:09 <scribe-jang> danc: we need things people can review and understand

14:09:14 <scribe-jang> art: yes, I think so...

14:09:45 <scribe-jang> danc: we could just put relaxng in the middle of the syntax spec...

14:09:56 <scribe-jang> maybe it _will_ communicate effectively with our target audience?

14:10:32 <scribe-jang> daveb: we ought to validate before we transform

14:10:40 <scribe-jang> ..then describe transforms for legal documents

14:10:44 <scribe-jang> danc: we don't

14:11:23 <scribe-jang> danc: we don't need to use a validation technology, we can just _tell_ people what is valid

14:11:35 <scribe-jang> ...we don't need to use a schema language to do that

14:11:44 <scribe-jang> jeremy: the current WD is quite intelligible

14:12:07 <scribe-jang> danc: current wd/sax-driven parsers are probably going to be quite popular, i think#

14:12:28 <scribe-jang> daveb: that's what jeremy and I have don in our parsers, trying to get WD to look like this

14:12:37 <scribe-jang> jeremy: what;s our audience?

14:13:06 <scribe-jang> agreed: with proviso that "as long as the primer covers the rest"...

14:13:19 <scribe-jang> the syntax doc is aimed at a target audience of parser producers

14:14:22 <scribe-jang> potential approach: have accurate description of syntax... non-normative relaxng appendix

14:14:33 <scribe-jang> we talk about transformations.

14:14:46 <scribe-jang> danc: I'm not sure the tranformation into canonical simple rdf can be done

14:14:54 <scribe-jang> because names for bnodes need introducing

14:15:17 <scribe-jang> jeremy: we need to extend rdf/xml a little to make this possible

14:15:23 <scribe-jang> the point is to go to ntriple in the end

14:15:33 <scribe-jang> the reason for staying within xml for as long as possible is...

14:15:53 <scribe-jang> because that's the natural setting for thinking about the abbreviated syntax _as_ abbreviations

14:16:17 <scribe-jang> danc: I think: everything we say is an abbreviation, specifying an xslt transofrm to eliminate it

14:16:23 <scribe-jang> abouteach comes up...

14:16:32 <scribe-jang> danc: I think it's still broken

14:16:49 <scribe-jang> pointer to unresolved test case to come (danc?)

14:17:02 <scribe-jang> art: re the transform into "simple" rdf/xml

14:17:18 <scribe-jang> jeremy: everywhere the uriref occurs, you have to allow a bnode name

14:17:33 <scribe-jang> jeremy: which you can write as _:name since that won't clash with rdf/xml

14:17:44 <scribe-jang> dave: can you describe snail in more detail?

14:17:48 <scribe-jang> jeremy...

14:17:57 <scribe-jang> you have a set of rules with a lhs and a rhs and some variables...

14:18:09 <scribe-jang> these rules apply only at the topmost elements of an rdf/xml document

14:18:15 <scribe-jang> ie, children of rdf:rdf

14:18:28 <scribe-jang> if an element matches the lhs then assign the variables according to lhs

14:18:31 <scribe-jang> and replace with rhs

14:18:36 <scribe-jang> and apply these rulkes in any order

14:18:40 <DanC> abouteach conundrum testcase: http://lists.w3.org/Archives/Public/www-rdf-comments/2001AprJun/0107.html

14:18:42 <scribe-jang> repeat until no matches

14:20:28 <scribe-jang> jeremy describes this in more detail

14:20:38 <scribe-jang> by staying at direct children of rdf:rdf

14:20:43 <scribe-jang> we avoid striping issues

14:20:59 <scribe-jang> we introduce bnode names as needed (as the transforms are applied)

14:21:18 <scribe-jang> you get deeper down by flattening

14:21:35 <scribe-jang> to flatten: you match a complex structure (with a propelt with a complex child)

14:22:04 <scribe-jang> jeremy is using complex xpaths to go several levels down

14:22:09 <scribe-jang> but always rooted at rdf:rdf

14:22:50 <scribe-jang> art: probably worth forwarding this to the list

14:22:59 <scribe-jang> jeremy: it's not very stable at the moment

14:23:06 <scribe-jang> all: never mind, get it on the web.

14:23:17 <scribe-jang> danc: xslt is turing complete

14:23:35 <scribe-jang> is jeremy's rules.xml less than turing-complete?

14:23:50 <scribe-jang> scribe-jang is now known as scribe-dajobe

14:23:57 <scribe-dajobe> jjc: not writing a prog language

14:24:21 <scribe-dajobe> dan: you have produced a reference impl?

14:24:43 <scribe-dajobe> jeremy: my view is what's important is what spec says, not ref impl - best way to convey meaning

14:24:53 <scribe-dajobe> dan: is approach any way less than turning complete?

14:25:03 <scribe-dajobe> jjc: not really

14:25:22 <scribe-dajobe> jan: yes is T-complete, not sure it matters. Question is how clear is it

14:25:40 <scribe-dajobe> ... could write in Java but waht need to do is write rdf/xml, here n-triples and show how to go between them

14:25:48 <scribe-dajobe> ... jeremy's approach is meant to be clear

14:25:57 <scribe-dajobe> dan: rules.xml meant to be opaque?

14:26:12 <scribe-dajobe> jeremy: will be compled into HTMl then will be clear patterns, visually appealing

14:26:28 <scribe-dajobe> such as propertyAttr meaing is given as abbr of propertyElt

14:26:39 <scribe-dajobe> and see set of abbreviated syntax as set of abbreviations

14:26:57 <scribe-dajobe> dan: xduce

14:27:15 <scribe-dajobe> dan: xsl is not compositional

14:27:39 <scribe-dajobe> jeremy: yes but ugly

14:28:37 <scribe-dajobe> ... goal is to be very clear in why its doing what its doing without

14:28:44 <scribe-dajobe> ... so that each step is small

14:28:54 <scribe-dajobe> ... for clarity of exposition

14:29:04 <scribe-dajobe> danc: bnodes?

14:29:19 <scribe-dajobe> jeremey: I use _:name

14:29:23 <scribe-dajobe> dan: so that is rdf-extended?

14:29:27 <scribe-dajobe> jjc: yes

14:29:47 <scribe-dajobe> ... and also need bagId and rdf:li counting, not sure best way to do latter

14:30:11 <scribe-dajobe> dajobe: rdf/xml extended required?

14:30:21 <scribe-dajobe> jan: yes, since need those bnode identifiers etc.

14:31:31 <scribe-dajobe> phayes arrives

14:31:36 <scribe-dajobe> transofmational grammar?

14:31:40 <scribe-dajobe> says jjc

14:31:50 <scribe-dajobe> dan: where do the rules become clear?

14:31:57 <scribe-dajobe> danbri: me too

14:32:21 <scribe-dajobe> s/danbri/dajobe/

14:32:29 <scribe-dajobe> ... it is the generated HTMl that is meant to do this?

14:32:49 <scribe-dajobe> jjc: the rules need to be expreessed in a way that captures their underlying simplicity

14:33:18 <scribe-dajobe> dajobe: right approach?

14:33:36 <scribe-dajobe> jan: seems attractive although concerned that you would be able to demonstrate

14:33:45 <scribe-dajobe> ... using rules in any order gives same end point

14:33:49 <jjc>http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Sep/att-0224/01-Transforming_RDF.html

14:33:52 <scribe-dajobe> ... and targetting parser producers

14:34:38 <scribe-dajobe> dajobe: better grammars and bnfs?

14:34:45 <scribe-dajobe> jjc: plenty of ones in arp, other places?

14:34:54 <scribe-dajobe> art: not formal

14:35:02 <scribe-dajobe> jjc: no method for formally describing triples at present

14:35:18 <scribe-dajobe> ... set of matching abbreviating syntax as set of abbreviations

14:35:25 <scribe-dajobe> danc: wrote xslt

14:35:37 <scribe-dajobe> jeremy: but that's a program

14:35:44 <scribe-dajobe> ... and programs have bugs

14:36:00 <scribe-dajobe> dan: but rules are also a program?

14:36:23 <scribe-dajobe> jeremy: yes, but in reducing description to some consise description

14:36:32 <scribe-dajobe> jan: scrutinisabily

14:37:13 <scribe-dajobe> discussion of URL above

14:39:03 <scribe-dajobe> jeremy: snaile meant to be a small language like prolog terms and variables, assign by matching and replace input by output

14:39:25 <scribe-dajobe> phayes: scope of variables

14:39:30 <scribe-dajobe> dajobe: syntax doc

14:39:34 <scribe-dajobe> dan: how complete?

14:39:50 <scribe-dajobe> jeremy: illustrative, not complete, probabbly twice as big in final.

14:39:55 <scribe-dajobe> ... reification is a pig

14:40:38 <scribe-dajobe> danc: wrote an rdf parser xslt in april ...

14:40:47 <scribe-dajobe> ... generated lots of issues

14:40:54 <scribe-dajobe> ... just changed to write ntriples

14:41:08 <scribe-dajobe> jeremy: had a quick look - what coverage of grammar?

14:41:37 <scribe-dajobe> danc: unclear, whatever I did in afternoon

14:42:05 <scribe-dajobe> jan: rdf/xml is a straighforward syntax with a billion corner cases, easy to do big bits but how do we complete it

14:42:35 <scribe-dajobe> danc: whatever we choose, if it turns out bad, we can change things

14:42:51 <scribe-dajobe> ... xslt cont.: no difference ot using another programming language

14:43:03 <scribe-dajobe> ... attribute grammar: Brian

14:43:18 <scribe-dajobe> ... kinda long to do but idioms are rather straightforward

14:43:44 <scribe-dajobe> jan: parser writers likely to have heard of attribute grammars

14:43:50 <scribe-dajobe> artb: brians work?

14:44:01 <scribe-dajobe> jeremy: relative complete

14:44:12 <scribe-dajobe> and spits out triples

14:44:31 <scribe-dajobe> ... might have other issues

14:45:29 <scribe-dajobe> dajobe: I'll try to work on EBNF sax-event like thing

14:45:53 <scribe-dajobe> artb: followup to earlier work?

14:46:01 <scribe-dajobe> jjc: I have an xml version

14:47:45 <scribe-dajobe> danc, jeremy: defitive document is in html with a lot of attachements

14:48:03 <scribe-dajobe> jeremey: may be the html generated form non-normative

14:49:19 <scribe-dajobe> we observe that the html will be normative and any non-normative xml appendix of mapping etc. is a bonus (excuse wording)

14:50:23 <scribe-dajobe> jan: decision is to plough on with various approaches?

14:50:37 <DanC> which approaches do folks expect to continue with?

14:50:47 <DanC> I don't gather my XSLT approach is all that interesting.

14:50:56 <scribe-dajobe> agenda item 5 outstanding issues

14:51:01 <jjc> I will be continue with snail

14:51:19 * scribe-dajobe will continue with mine bnf-y

14:52:21 <scribe-dajobe> daveb runs fdown the issue list

14:52:26 <scribe-dajobe> syntax unclear

14:52:37 <scribe-dajobe> quoting, reification can't deal with yet

14:52:40 <scribe-dajobe> bagid

14:52:43 <scribe-dajobe> nested bagids

14:52:47 <scribe-dajobe> qnames can't represent uris

14:53:08 <scribe-dajobe> rdfms-quoting

14:53:24 <ArtB>http://www.w3.org/2000/03/rdf-tracking/

14:53:34 <scribe-dajobe> rdfms-qnames-can't-represent-all-uris

14:53:39 <scribe-dajobe> rdfms-replace-value

14:53:52 <scribe-dajobe> rdfms-qname-uri-mapping

14:54:22 <DanC> I don't see "M&S" in the TOC

14:54:25 <scribe-dajobe> the general idea: that issues should move where appropraite to be attached to the document

14:54:29 <scribe-dajobe> half go to MT

14:54:34 <scribe-dajobe> rougly half go to syntax

14:54:38 <scribe-dajobe> and there are a few left over

14:54:46 <scribe-dajobe> and teh editor of a document looks after those issues

14:54:50 <DanC> ah... found it. pls fix the TOC

14:55:15 <scribe-dajobe> jeremy: a number of these (even syntax issues) are bluesky

14:55:23 <scribe-dajobe> it would be helpful maybe to triage?

14:55:41 <scribe-dajobe> eg: qnames as attributes: these kinds of things we consider out of scope

14:55:55 <scribe-dajobe> dave: we have to address these

14:56:12 <scribe-dajobe> even if it is: this is an extension, it is out of the scope of the current wg

14:56:51 <scribe-dajobe> the syntax document editor handles the resolution of these issues

14:56:58 <scribe-dajobe> since they're synhtax related

14:57:43 * DanC was distracted by a call on the other line

14:57:59 <scribe-dajobe> and the doc editors/syntax people would write the text

14:58:05 <scribe-dajobe> the idea is simply to speed up the work

14:58:50 <scribe-dajobe> we'll have to take on board i18n issues too (jeremy)

14:59:02 <scribe-dajobe> jeremy: homework?

14:59:49 <scribe-dajobe> rdfms-syntax-desc-clarity

15:00:01 <scribe-dajobe> rdfms-resource-semantics: this is model

15:00:16 <DanC> ============= triage of M&S issues.

15:00:21 <scribe-dajobe> equiv uris are mt

15:00:30 <DanC> 1. rdfms-resource-semantics: not syntax

15:00:47 <DanC> 2. rdf-equivalent-uri's: not syntax. model

15:00:50 <scribe-dajobe> fragments: sort of modelish, unclear

15:01:06 <scribe-dajobe> we're looking for obvious syntax

15:01:06 <DanC> 3. rdfms-fragments: not syntax

15:01:19 <scribe-dajobe> literalsubjects, model

15:01:32 <scribe-dajobe> xmllang model?

15:03:16 * rdfms-literalsubjects: not syntax

15:03:30 * rdfms-xmllang: not syntax

15:03:42 * rdfms-syntax-desc-clarity: SYNTAX

15:03:51 * rdfms-contexts. not syn

15:04:52 * rdfms-identity-of-statements: not to be decided in the syntax doc, might show up in the syntax doc

15:05:11 * rdf-containers-otherapproaches: not syntax

15:06:16 * rdf-equivalent-representations: not syntax

15:06:27 * rdf-formal-semantics: not syntax

15:06:38 * rdfms-quoting: SYNTAX

15:06:48 * rdfms-nested-bagIDs: SYNTAX

15:07:01 * rdfms-qnames-can't represent-all-uris: SYNTAX

15:07:10 * rdfms-replace-value: SYNTAX

15:07:19 * rdfms-propElt-id-with-dr: SYNTAX

15:07:30 * rdfms-formal-grammar: SYNTAX

15:07:50 * rdfms-seq-representation: not syntax

15:08:01 * rdfms-xml-literal-namespaces: SYNTAX

15:08:30 * rdfms-assertion, * rdfms-logical-formalism * rdfms-logical-terminology: not syntax (though may show up)

15:08:46 * rdfms-qname-uri-mapping: SYNTAX

15:08:56 * rdfms-boolean-valued-properties: not syntax

15:09:06 * rdfms-qnames-as-attrib-values: SYNTAX

15:09:10 <DanC> (no, btw)

15:09:20 * rdfms-validating-embedded-rdf: SYNTAX

15:10:10 * rdfms-xml-base, * mime-types-for-rdf-docs, * rdfms-abouteach, * rdfms-reification-required, * rdfms-rdf-names-use: SYNTAX

15:10:19 * rdfms-editorial: whatever

15:11:32 <scribe-dajobe> action: daveb to take a first pass at these identified issues

15:11:33 <DanC> did somebody just volunteer to do a bunch of test cases?

15:11:58 <scribe-dajobe> test cases where appropriate (ie, for in scope issues)

15:12:48 <scribe-dajobe> jeremy: do we write up the closed issues in an appendix too?

15:12:55 <scribe-dajobe> daveb: I've actually done this in the wd

15:13:18 <scribe-dajobe> i expect newer versions will say "we're addressing these issues, see section X"

15:13:46 <scribe-dajobe> AOB?

15:14:01 <scribe-dajobe> danc: when do we want to do the next draft (for publishing)

15:14:11 <scribe-dajobe> dave: asap, but we're a bit busy here at the moment

15:14:19 <scribe-dajobe> want to get some mapping into the next version

15:14:37 <scribe-dajobe> with words about how you would use that

15:14:54 <scribe-dajobe> we'd all like to finish by christmas (AGREED)

15:15:32 <scribe-dajobe> danc: and it'd be good to edit the syntax spec and making test cases, and approve those changes in the telecon

15:15:39 <scribe-dajobe> rather than resolve to make changes

15:15:57 <scribe-dajobe> daveb: so the next version of doc with attached issues "TODO"

15:16:13 <scribe-dajobe> danc: the list we just made is an appendix with pointers into the document

15:16:27 <scribe-dajobe> jeremy: a lot of these issues come from rdf/xml -> triples

15:17:11 <scribe-dajobe> danc: it's cost-effectiveness: incluing explanations takes time, but avoids FAQs

15:17:52 <scribe-dajobe> the test cases get listed in the test case document... and on the web, the syntax doc points to test cases

15:18:03 <scribe-dajobe> are we stalling here?

15:18:11 <scribe-dajobe> ok, we've got a list of issues

15:18:26 <scribe-dajobe> we'll take this to email (rdfcore)

15:18:35 <scribe-dajobe> mjeeting closed

15:19:01 <scribe-dajobe> not-id-and-resource-attr SYNTAX

15:19:04 <scribe-dajobe> (we missed this)

15:19:18 <scribe-dajobe> rdfms-difference-between-ID-and-about: SYNTAX