WoT Discovery

F2F minutes

vF2F day 1

vF2F day 2

vF2F day 3

vF2F day 4

vF2F day 5

vF2F day 6

Kaz: will send a reminder to the IEC guys and ask them for slides

McCool: "will give feedback for T2TRG"
… don't see any issues, can accept?

(no objection)

day 1 minutes accepted and to be marked as "reviewed"

Kaz: note day 2 is the Discovery day

McCool: don't see major problems; any problems?

(none)

McCool: mark as "reviewed"
… official acceptance during the main call later

Issue 145

<glomb> https://github.com/w3c/wot-discovery/pull/145 PR 145: Listing with chunked transfer

PR not issue

Farshid presents

Three proposal: pagination / headers / chunking

Chunking: Streaming instead of pagination

<kaz> Preview: 6.2.2.1.5 Listing

mm issue:JSON-LD needs context, Farshid: not necessary

HTTP 1.1 chunked transfer encoding

Widely supported

HTTP/2 -> http frames

Where is it applied?

McCool: browser support it, variable chunk size

McCool: almost invisible

clients have to request content-lenght

Question: how big is the entire message?

McCool: Content-length optional

Farshid: expensive to calculate, can be optimized

McCool: client can do head

Farshid: some concerns that it is not enough, does not prevent pagination

Christian: want to have both options

<kaz> Linked Data Platform Paging 1.0: 4.2 Simple paging flow using redirects

Both mandatory: both optional?

Farshid: Query API should provide pagination

McCool: Server should optimize

McCool: extra work to realize

Chunking for large TDs

Payload body: array of TDs

Christian: want possibility to add extra info

McCool: clean up draft, chunking as basic functionality

McCool: leave pagination for the moment and ask for review

Farshid: not compatible

Chunking w/ current draft

Farshid: Header based approach can be added to chunking

<kaz> PR 130 Preview: 6.2.2.1.5 Listing

<kaz> proposal: merge PR 145; add a new PR based on PR 130

https://github.com/w3c/wot-discovery/pull/145 should be merged. https://github.com/w3c/wot-discovery/pull/130.

130 will be closed

Cristiano: +1 for merging this PR
... Then cleanup
... What about extra info in the body?
... Issue Mix of JSON + JSON-LD
... Chunking also applies for submitting TDs
... Large TDs vs. large sets of TDs.

McCool: Pagination info in body or header: to be discussed.

Editorial PRs

<kaz> PR 136: Validation error example

McCool: Editorial PRs should be merged

<kaz> PR 137: Update jsonpath draft URL and date

<kaz> PR 138: Improve directory intro

Issue 147

Issue 147: Add reference to Discovery paper by Arne Broering, Soumya Kanti Datta, and Christian Bonnet

Fundamental work should be acknowledged

McCool: Some more work to be cited?

<kaz> [5min break]

<kaz> (Christian and Soumya leaves)

<kaz> (Ben joins)

Issue 133

<kaz> Issue 133 - Refactoring TDD Thing Description

Farshid: please speak about the refactored td

Ben: single things properties, combining listing and filtering
… spitting in three events
… anyother changes?

Farshid: td -> is now things. More readable

<kaz> i/please peak/topic: Issue 133/

Farshid: is it ok to call them things
… ?

McCool: I agree with the changes
… it is a reasonable change, although it is less readable
… is a registration an action?

Farshid: yep

McCool: another issue property names in url is not consistent
… should we create a convention for urls?
… it is a whole other discussion
… urls can change but not the affordance names

Farshid: it is a another issue... it means that the URL can be changed in the spec?

McCool: maybe a TM can help

Farshid: should we call things or tds?

McCool: vote for td
… @type : "Thing" is a legacy definition
… I am ok eitherway
… just stick with things
… is sparql a property
… is weird

Cristiano: did we solve the problem of choosing which form to use?

Farshid: sparql is special

Ben: we cannot add sparql as query filter
… property vs action is an open question

Andrea: why sparql is not a property?

Ben: we cannot describe it correctly as a form of things

McCool: jsonpath would be much more common
… sparql is even optional

Ben: sparql is separate because jsonpath and xpath is more a fragment than a proper high level query lang
… also we have still to decide what is really normative

McCool: I would say that once we have a thingmodel we'll use it

(usually TMs does not have fixed urls)

McCool: closing the meeting, let's keep discuss on the issue

<kaz> [adjourned]