See also: IRC log
<trackbot> Date: 21 January 2015
<danbri> morning Gregg!
<danbri> Issues flagged for discussion: https://github.com/w3c/csvw/issues?q=is%3Aopen+is%3Aissue+label%3A%22Requires+telcon+discussion%2Fdecision%22
<danbri> standing agenda: https://www.w3.org/2013/csvw/wiki/Meetings#Standing_Agenda
<danbri> I suggest process as follows. For each editor who joins, we get them to flag issues worthy of discussion, then jump in.
Morning Dan!
<scribe> scribenick: gkellogg
ivan: the two conversion docs are pending metadata issues.
… gregg did an experimental writeup last weekend, but it still depends on many mentadata issues.
jtandy: one issue is how to determine the identifier in the metadata document describes the metadata for a table, or identifies the table itself.
<danbri> is this the gregg msg ivan referred to? http://lists.w3.org/Archives/Public/public-csv-wg/2015Jan/0042.html (or something in github)
… It makes a difference in the conversion docs.
ivan: at the moment, the @id is used as the id of a graph.
… The issue is clear to me; do we need a separate ID fo the metadata?
<danbri> gkellogg: "… it might be useful to distinguish table's id, … or other metadata elements. Not entirely clear what happens. In my draft I suggested the id of the table group might be used as id of the table group in the output.
<danbri> … main thing is to be consistent.
<danbri> I suspect we'll receive some comments about conflation here, but Jeremy's attempt to add a fragid seems to reconcile that
<danbri> I was running across a couple issues trying to do some rdfish things with the output, but that's not really one of our requirements.
<danbri> ivan: that means this is one of those issues we can settle on without jeni?
<danbri> gregg: we might need her for all of them :)
ivan: that means that we need JeniT, no?
<danbri> https://github.com/w3c/csvw/issues/106
<danbri> … not this one
<danbri> meta issue: we can't find what we're looking for within github. Too many!
https://github.com/w3c/csvw/issues/113
https://github.com/w3c/csvw/issues/93
<danbri> gregg: i tried to do a little DOAP description in CSV and … I couldn't
<danbri> … we talked about way of changing the subjecgt of the generated triples
<danbri> … there is a url template for a row
<danbri> presumably you create a url templatefor a row where all rows have same subject
<danbri> … so maybe these things need to be in a list
<danbri> … otherwise all get smushed together.
<danbri> jtandy: back at f2f we talked about adding an attr called subject url, ...
jtanday: we talked about adding “subjectUrl” in the F2F.
… I didn’t include in the scope of the conversion document, as it was a bit difficult.
… It was so you could use a URI template to determine the subject of the output triples for a given row.
… At the moment, all triples with the same row have to have the same subject, so that we can create more than one resource per row.
… This was being discussed as people were getting coffee.
ivan: doesn’t the current document say that we have a urlTemplate with the schema, and that’s used to generate the subject for a row.
jtanday: gkellogg’s suggesting doing it per column.
ivan: my recollection was that we discussed it and decided it goes too far.
danbri: when I worked through some use cases, the problem with the simple model touched on three different things, and I found myself getting beyond those proposals.
<danbri> here is my Event example, https://github.com/w3c/csvw/tree/gh-pages/examples/tests/scenarios/events/output
<danbri> (it corresponds to actual json-ld consumed at Google)
<danbri> expected triples here: https://github.com/w3c/csvw/blob/gh-pages/examples/tests/scenarios/events/output/expected-triples.txt
<danbri> csv: https://github.com/w3c/csvw/blob/gh-pages/examples/tests/scenarios/events/source/events-listing.csv
<danbri> gkellogg: looking, … I don't see how columns can vary in current proposal
A URI template property that may be used to create a unique identifier for each row when mapping data to other formats.
<danbri> ivan: at moment you can have url template on schema, or on a col
ivan: I know we needed something for a column where the object is a URI and not a LIteral
<jtandy> http://w3c.github.io/csvw/csv2rdf/
<danbri> hi Davide
jtandy: in the conversion document, urlTemplate is used for creating a URL object, not subject.
<DavideCeolin> hi all
ivan: we agreed that predicateUrl was not a template.
<danbri> gregg: re relation to metadata doc, a number of places that trip us up
<danbri> around use of json-ld … and some other interpretations, how the metadata docs are resolved, what the language implied for embedded metadata is, etc.
<danbri> for those i think we need JeniT to make progress
ivan: someone was trying to convince us that the rows should use :_n, or should be in a list.
<danbri> ivan: ther's an issue I'm looking for, … someone sent in an issue end December 2014, trying to convince us for the RDF output, the rows should use rdf_n collections, or lists
<jtandy> https://github.com/w3c/csvw/issues/116
<jtandy> (this was the issue about use of blank node shorthand)
danbri: schema.org had a bunch of issues crop up, which indicate that it may take forever to solve things.
… would it help to add milestones.
<jtandy> (sorry dropped out of call)
<danbri> gkellogg: let's consider which are blocking others
<danbri> e.g. metadata merging
<danbri> …some other rules we might use e.g. normative text around processing of input rows
<danbri> i tried this but JeniT pointed out that was being done in a non-normative section
<danbri> https://github.com/w3c/csvw/issues/116
ivan: I realized that the RDF generation doesn’t add a row number to the output; that may be an oversite.
… There’s no trace of the order of the original rows.
… you may want to find out the row order.
ivan: the user can do it with a template, but the author needs to add some steps.
jtandy: it occurs to me that ordering may be important to some people; I thought it might make sense to add in something like a dc:source attribute, or isDefinedBy to refer to the fragment identifier of the source in the row.
ivan: the problem with using fragment identifiers is conflating the datastructure with the identifiers.
jtandy: we could allow a mechanism to use an RDF:List optionally.
… often order is irrelevant.
danbri: can also push the problem into conversion applications.
<danbri> (some CSVs will be consistent across rows, allowing you to merge the whole lot into a giant graph. Others will have different prov for each row, and may get garbage if all rows are merged. we should push this problem into tools/apps for now imho)
jtandy: we had discussed the ability to add virtual columns using the metadata description. Let’s say we wanted to add the source for every row, but it wasn’t actually there, we could create a “virtual” column with an _col magic variable.
ivan: that would probably screw up my implementations.
… The use of rdf:List is a “sledge hammer”.
<jtandy> (sorry - meant the magic variable '_row', as defined in http://w3c.github.io/csvw/metadata/index.html as a URI template property
… The simplest thing is to do the obvious one, and generate an extra triple containing the row number.
… Possibly using some option to control this.
<danbri> danbri: conclusions? next 10 mins?
<danbri> gkellogg: we talked thru this but didn't resolve
<danbri> we may have an issue w.r.t. not using json-ld too heavily
<danbri> … could back off to saying values of common properties are values and uris, not deeper structures
<danbri> ivan: fully agree
<danbri> gregg: an intermediate node may need a json-ld processor
{“dc:contributor”: {“@id: “@danbry”, “foaf:name”: “Dan”}}
<danbri> gkellogg: at least for timebeing just have an enumerated set of common property structures, without too much depth
https://github.com/w3c/csvw/issues/142
<danbri> proposal is that we limit the values that can be used on common properties to be one of those 6 options
<danbri> ivan: why need that 2nd one?
<danbri> "dc:description": {"@value": "String value with no language"},
<danbri> t-3
jtandy: I’d expect us to have to handle graphs of triples when dealing with note annotations.
gkellogg: notes might not have the same restrictions as common properties.
<danbri> adjourned.
<danbri> which bot to ask for minutes?
<danbri> ivanbot?
<ivan> trackbot, end telcon