CSV on the Web Working Group Teleconference

23 Apr 2014


See also: IRC log


Andy Seaborne (AndyS), Ivan Herman (ivan), Jeni Tennison (JeniT), Jeremy Tandy (jtandy), David Ceolin (DavideCeolin), Eric Stephan (ericstephan), Jürgen Umbrich (jumbrich), Anastasia Dimou (andimou), Rufus Pollock (rgrp, IRC only)
Phil, Stasinos, Davide, Yakov, DanBri


approving minutes


RESOLUTION: minutes from http://www.w3.org/2014/04/09-csvw-minutes.html are approved

Use Cases & Requirements

jtandy: I've seen a few additions coming in with new use cases
... eg HL7
... also a couple of others to dig into
... the major piece of work is to go through the requirements & identify the ones we want to approve
... we should also flesh out the requirements
... which are currently stubs

jtandy: I'm hoping to find some time later this week to get that started

ivan: there are examples that Yakov collected re internationalisation
... it would be good to add those to the use case document too

jtandy: I recall those conversations, and agree we should include them, they're not there currently

JeniT: how should we keep track of the new additions? shall we use GitHub issues?

jtandy: yes, good idea, we'll raise new use cases there

ivan: who should put in the first set of issues to github?
... I'll add the internationalisation one

jenit: I'll add the HL7 one
... jtandy, could you do an initial split on requirements so we have something to argue about?

jtandy: will do
... regrets for next week
... but Davide/Eric could talk through the candidate list for next week's meeting?

ericstephan, DavideCeolin: sounds good

jtandy: I think the majority of requirements are good, there's possibly a small number that are duplicates

DavideCeolin: I can start doing that partitioning


AndyS: Gregg's put in some materials discussing templating RDF outputs
... that should be a good framework for designing the process
... people can add their own templates
... but we can also define annotations in terms of the templates
... which can then be overwritten
... the only issues I can see are that to have a template you have to have a syntax

AndyS: turtle is fairly natural
... and we should probably pick that
... we'd all appreciate discussion & feedback on the list
... we're not going to resolve the issue about how complicated to make that
... the balance between having some level of RDF & how complicated to make the remapping
... it overlaps with other toolchains
... but you don't want to learn lots of different tool to get RDF out of CSV files

ivan: how generic are the templates? what if I want to generate JSON or XML?
... are the templates very RDF specific?

AndyS: I don't know; it works easily in RDF because you're producing triples
... whereas JSON is about generating the tree
... it would be interesting to work through that
... there's some level of JSON generation within a fixed framework
... more complicated arrangements where the top-level wrapping isn't an array then it could be complex
... Gregg is probably the person to answer that, please could you bring it up on the list?
... on the mailing list for wide comment

ivan: I was asking because if the templates are generic then this has an influence on syntax
... because you don't want turtle because anti-RDF people won't use it

AndyS: there is that: you have to say what you want generated anyway
... to some extent, if there's a template system for JSON and one for RDF triples
... then that's probably ok
... I'm still not keen on using JSON-LD as the templating system for RDF
... and going down the road where JSON-LD is acceptable for JSON output

ivan: I don't know, I was thinking about the template in a JSON format

AndyS: I don't think these are technical issues, more external requirements issues
... and that's why I want more input from the wider group


ivan: it would be good to mark which of the documents are the ones you want us to comment on/look at

<rgrp> JeniT: good points

JeniT: please could everyone comment on overall direction on the mailing list
... also: is it reusing SPARQL CONSTRUCT?

AndyS: no, it's a syntax-based template based on turtle

<scribe> ACTION: Jeni to start discussion about level of mapping [recorded in http://www.w3.org/2014/04/23-csvw-minutes.html#action01]

<trackbot> Created ACTION-13 - Start discussion about level of mapping [on Jeni Tennison - due 2014-04-30].



jenit: there are still issues here
... particularly around whether we support metadata embedded within CSV files
... I'll frame this on the mailing list and then try to have a discussion on the call next week or the week after

Metadata Vocabulary

jenit: after call last time, I got in touch with rufus re editing spec on metadata vocabulary, based on data package work
... he will edit that document, with me & possibly Ross Jones as editors
... we'll aim for a first draft in two weeks

<rgrp> ok

<rgrp> good to have a deadline :-)


jtandy: just to confirm, the datapackage stuff exists
... AndyS, Gregg & myself are looking at templating
... are we looking at merging those

<rgrp> ivan: i'm rgrp on github

<AndyS> I'd assumed two docs - the packaging was the format, then RDF/JSON mapping used that. But one is OK.

jenit: yes, we'll get to a coherent state where there's a metadata vocabulary and the RDF/JSON mapping documents reference or use that

ivan: the packaging format isn't the metadata vocabulary, right?
... the vocabulary is another issue that we haven't touched at
... are the template for RDF part of the metadata or not

<jtandy> JeniT: there are three things -

<jtandy> ... the packaging of metadata and multiple csv files together

<jtandy> ... the metadata vocabulary for describing the CSV content

<jtandy> ... the mapping of CSV content to RDF, JSON etc.

<jtandy> ... if the latter two are combined then this will be a design decision

<rgrp> i'd have to say whilst i think 2 + 3 are clearly strongly related they shoudl be distinct ...

<jtandy> ... the OKF data protocol work merges these requirements at present; wanting to separate these out

<jtandy> no probs

<rgrp> jtandy: i think (3) isn't covered by data protocols stuff and 1+2 are separated to the extent that data package spec is separate from json table schema (+ these 2 get combined to make tabular data pacakge)

ivan: for our own planning, will we go through another publication round in 2-3 weeks?

<rgrp> jtandy: good points

jenit: I think closer to 4 weeks

<jtandy> rgrp: thanks for clarification

Summary of Action Items

[NEW] ACTION: Jeni to start discussion about level of mapping [recorded in http://www.w3.org/2014/04/23-csvw-minutes.html#action01]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.138 (CVS log)
$Date: 2014-04-23 12:48:32 $