W3C


Web Annotation Working Group Teleconference

08 Jul 2015

Agenda

See also: IRC log

Attendees

Present
Frederick Hirsch, Rob Sanderson, Ray Denenberg, TB Dinesh, Jacob Jett, Ben De Meester, Tim Cole, Matt Haas, Doug Schepers (shepazu), Ivan Herman, Benjamin Young, Chris Birk (chrisbirk), Davis Salisbury, Paolo Ciccarese, Kyrce Swenson
Regrets
Chair
Frederick_Hirsch, Rob_Sanderson
Scribe
Jacob

Contents


<trackbot> Date: 08 July 2015

Agenda Review, Scribe Selection, Announcements

<RayD> ray

<fjh> ScribeNick: Jacob

<fjh> Web Annotation Protocol FPWD published, https://lists.w3.org/Archives/Public/public-annotation/2015Jul/0023.html

fjh: Add tweaks / clarifications to the agenda

Minutes Approval

<fjh> proposed RESOLUTION: Minutes from 24 June approved, http://www.w3.org/2015/06/24-annotation-minutes.html

RESOLUTION: Minutes from 24 June approved, http://www.w3.org/2015/06/24-annotation-minutes.html

TPAC schedule

<fjh> doug: no need to change schedule

doug: no change to schedule, re: ivan had conflict

<fjh> TPAC Annotation schedule unchanged, Mon/Tue

doug: [doug] attending web anno but popping out to help with other working groups as needed

<fjh> no change to Annotation schedule

Testing - Sparql based model validation

<fjh> https://lists.w3.org/Archives/Public/public-annotation/2015Jun/0275.html

<fjh> fjh: is there an action here for someone?

fjh: is someone taking an action to actually do the testing that Rob suggested?

<chrisbirk> I can take that

<TimCole> here's the list of SPARQL queries used by Open Annotation for Community Group work

<TimCole> https://github.com/uq-eresearch/lorestore/blob/master/src/main/resources/OAConstraintsSPARQL.json

azaroth: should test feasibility and report back to the group with a summary of how it works, if it fufill its requirements, and what it would take to update moving forward

<PaoloCiccarese> q

chrisbirk: can take this action item

TimCole: pasted in link to the Anna's list of SPARQL that were used for testing the OA community specs
... useful starting place, could examine if they can be used to suggest the form of protocol testing, happy to help with that action

<davis_salisbury> Davis (me) has called in

<ivan> Davis

<Kyrce> I am muted on my phone, I am not muted on webex.

<Zakim> azaroth, you wanted to offer help, & re protocol testing

fjh: confirming the Chris volunteered to conduct that testing, Tim volunteered to help

azaroth: also volunteering to help; familiar with Anna's queries, can confirm which ones still apply and which need changing

<fjh> tasks - confirm that query approach can work, review exisitng queries for needed change, share draft of test results

azaroth: mentioned protocol testing, once queries have been examined, then protocol testing (for HTTP conformance) can be approached separately
... once model is testable, is much easier, can examine whether or not headers, status codes are correct

<azaroth> +1 to not testing protocol yet

<Zakim> fjh, you wanted to mention protocol

fjh: may be premature to test the protocol until collections and similar issues are resolved

<azaroth> +1 to Paolo, re testing the /model/ vs testing the /serialization/ [which we can also do]

PaoloCiccarese: regarding queries, even if the rdf structure is changed, recompiling the annos out of the triple store, they can be compiled different ways, testing queries is helpful because they ignore the order (structure) of the rdf, however, it is a major task
... in addition to this kind of testing, we need to consider framing and how to test framing, may be difficult to validate

<azaroth> http://json-ld.org/spec/latest/json-ld-framing/

PaoloCiccarese: framing -- precise json-ld structure, (because serializations can be different)

<azaroth> Sample annotation frame: https://github.com/IIIF/iiif.io/blob/master/source/api/presentation/2/annotation_frame.json

PaoloCiccarese: annotation parts can be put into different places in the serializations, so client-side is complicated because a annotation can have many possible permutations of serializations

<Zakim> azaroth, you wanted to distinguish application level validation, vs testing requirements for implementation conformance

PaoloCiccarese: SPARQL always works, but other forms of validation will be much more complex

azaroth: verbally +1 distinguishing testing the model from testing the serialization
... also application vs. incoming requests, etc. [lost some of the permutations Rob discussed]

<Zakim> fjh, you wanted to ask if we need annotation canonical serialization

fjh: will we have to define serialization canonicalization and test this, will it be complicated?

azaroth: simpler in our case because can test json keys and bits and pieces, similar to past epub work

<azaroth> EPub oriented json-schema for OA: http://www.idpf.org/epub/oa/schema/oa-epub-schema.json

ivan: framing document -- general framing algorithm was quite complicated and community group could not come up with a satisfactory algorithm that could be tested and meet the requirements
... important to understand where the problems are when framing is defined

<ivan> http://w3c.github.io/csvw/csv2json/#nesting-objects

ivan: useful to have a target to help define the shape of the serializations, one example of such a thing is csvw solution

PaoloCiccarese: using framing for a while, normal framing works pretty well, using framing from CG to avoid creating own, easy if structure is simple, but becomes complicated when annos are structured
... because of the SPARQL queries, was able to implement a two-level strategy for validation, one syntactic validation and one semantic validation

TimCole: first, these are two separate issues, need way to validate annos to assure that docs are proper, however, most folks will not want to deal with rdf and so we'll accommodate that
... one solution could be to develop some services that offload some of the validation work
... need to face up to the fact that most folks won't add validation to their clients

shepazu: if this is the case, is there another way to test the model
... problem seems specific to semantic web and not data models
... specific entities (structures) might be necessary but not in a specific order
... don't need to spend the energy on an rdf frame

fjh: need to clarify what our goal is and what we're going to do
... Chris, Tim and Rob going to work on the testing of the model

azaroth: do we want to track testing related issues in the current git repo or do want a new/separate one?

<chrisbirk> I think a separate would be easiest

ivan: no set rule, in the csv group, we just have a single repo, but if it is easier to manage...
... whoever is managing should give their preference

fjh: moving further discussion offline

<fjh> +1 whoever manages tests should arrange tracking

fjh: what is the time estimate for this activity

chrisbirk: sometime in the next week or so

Model: Open vs Web Annotation namespace etc

<fjh> https://lists.w3.org/Archives/Public/public-annotation/2015Jul/0011.html

fjh: seems to be general agreement on the list to go with the web anno namespace
... can expect additional changes in the future but no reason not to move to the namespace now

azaroth: naming change ok, best practice is not to change namespace unless absolutely necessary, personal preference not to change namespace
... should consider the namespace issue separately from the naming issue

PaoloCiccarese: is web anno going to supersede the open anno name / namespace; what is the public message to the folks who have already implemented older versions of the oa model?
... are we saying that open anno is dead?

azaroth: don't want to change the namespace if we change to something that is not 100% back compatible
... market will decide on which model will be used

ivan: problem with namespace issue, in practice, is back compatibility, old foaf files are still valid foaf files, same has been done with rdf and owl, those cases are clear agreements in the charters to focus on back compatibility
... web anno has no such agreement, the message is that -- web anno is a more advanced version of the older open anno doc/standard

<fjh> merit to Paolo's concern and Tim's rationale

TimCole: understand and sympathetic to that position, CG members also understood that the working group's output was going to be an evolution; so if we create a new namespace it seems like it could bifurcate the community

<fjh> existing implementations matter

TimCole: risk that some current implementers may see the new namespace as a competing standard rather than a new rendition of open anno

fjh: shouldn't change namespaces just as a "hygiene" issue, do we have a compelling reason to make the change?

<azaroth> http://www.openannotation.org/spec/core/

<ivan> there is probably no reason, fjh, to do it *now*

<fjh> I think we should reserve the right to change, but only if we need to, when we have good reasons. I don't hear that we are at that point now. Agree we can, but we should give it consideration later.

azaroth: final thought, CG did not publish a definitive document, output was a draft, updating the draft could mitigate some of the potential problems, so the namespace issue may not be a big deal

fjh: not ready to make a decision on this call, no compelling rush, can we continue this discussion on list?

azaroth: clarification, the documents should talk about web anno but not so with the namespace

RangeFinder

<shepazu> http://w3c.github.io/rangefinder/

shepazu: range finder draft, primary change based on feedback, may need different kinds of unicode folding for different languages
... changed from boolian value to typed value

<shepazu> http://w3c.github.io/rangefinder/#idl-def-UnicodeEquivalenceType

shepazu: for the UnicodeEquivalenceType

<fjh> ACTION: chrisbirk to propose model testing, review existing Sparql testing for Web Annotation Model (with Tim Cole and Rob) [recorded in http://www.w3.org/2015/07/08-annotation-minutes.html#action01]

<trackbot> Error finding 'chrisbirk'. You can review and register nicknames at <http://www.w3.org/annotation/track/users>.

shepazu: records which kind of equivalence is preferred, but not sure if browsers are going to be willing to something this complicated for unicode equivalence matching

<fjh> I think Doug is suggesting MUST for one Unicode type support

shepazu: would be surprised if they do more than one kind, so we should probably pick one, should see if the internationalization group can make progress on convincing the browser group to do different kinds of canonicalization
... if anyone has any feedback on the draft would be interested in it

<fjh> +1 to publish FPWD soon

ivan: can we publish this document as a fpwd?

shepazu: would like to make an additional round of editing to make the examples match the existing api, then would be ok to publish, so possibly by next week...

<azaroth> +1

<TimCole> +1

Protocol

ivan: publishing community is interested in this, so publishing the draft will be helpful

<ivan> +1 to azaroth

fjh: Ray's suggestion for naming the protocol document may be helpful for working around some of the issues on the list, seems like folks are talking past one another

azaroth: Erik's concerns seem deeper than that

fjh: missing why this is breaking the web, what is the problem?

azaroth: my understanding, architecturally, a generic http client should only be governed by the http spec, and the reason the web is successful is because there is no typed coupling at the app level betweem clients and servers using http

<fjh> don't media types create coupling?

azaroth: the protocol doc tries to constrain this by suggesting a simple header rather than spelling out every permutation [?]; Erik's concern is that http doesn't require that every uri handle POST
... if we agree with that position then we would have to have a protocol that requires creating annos using post, patch, etc. becausre http doesn't require it, but that seems to defeat the entire purpose of the anno standard
... would make testing impossible and conformance meaningless
... everything would be consume only, whereas we're trying to get beyond that

<fjh> we can ask the tag to look at our protocol spec regarding this concern

ivan: why is he happy with ldp but unhappy with what we have?

azaroth: could reuse some of the weasel words in ldp

Adjourn

<fjh> thanks Jacob for scribing

<ivan> trackbot, end telcon

Summary of Action Items

[NEW] ACTION: chrisbirk to propose model testing, review existing Sparql testing for Web Annotation Model (with Tim Cole and Rob) [recorded in http://www.w3.org/2015/07/08-annotation-minutes.html#action01]
 
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.140 (CVS log)
$Date: 2015/07/09 04:23:13 $