Meeting minutes
minutes
<kaz> Feb-8
no objections to publishing the minutes
Extending the discovery slot to 2h
McCool: We can try to finish important things in the first hour and continue with less important things in the second hour
McCool: Will do an extra hour today and discuss it again next week. Some people have conflicts.
F2F schedule
<kaz> F2F topics
McCool: The date is fixed
McCool: Need to start scheduling some topics
McCool: In particular, the discovery topics. E.g. review of current spec such as introduction, info. model, serialization
Cristiano: One open issue which we can discuss in the F2F is the framing.
Andrea: Cannot convert back and forth between JSON-LD and RDF
Christian: It makes sense to have TD people and also Victor Charpenay who designed the TD ontology.
<kaz> kaz: should add semantic search, etc.
McCool: Each day is 3h. If longer, we'll have to break the slot into 2 days.
<McCool> https://
PR Specify self-described exploration #115
<kaz> PR 115 - Specify self-described exploration
McCool: References in assertions cause issues in the generated output
McCool: What about filtering / selecting a part of TD, e.g. using JSONPath
<kaz> 6.1 Self-description
McCool: For example, to ask devices to respond only if their TDs have a particular property
McCool: no objections to merging.
Refine the information model #118
<kaz> PR 118 - Refine the information model
McCool: prefer the overview.
Farshid: Didn't wanted to show we have a new class
Farshid: This diagram makes it clear this is optional.
McCool: Do I need all information in the directory? It is all or nothing.
<FarshidT> https://
Farshid: The diagram tool is used in the other specs as well.
McCool: Little confusing diagram.
Farshid: yes, I can change that.
Andrea explaining some context of the diagram tool.
McCool: Guess, in the future, reopen the discussion. Okay with merging it for now.
McCool: Any comments on it?
Farshid Adds some comments about the diagram.
Farshid: Check terminology.
McCool: I take a look
McCool: probably want to be more specific.
Farshid: When I cite the terminology I am more specific.
McCool: question is, if the terminology of partial TD consistent.
McCool adds issue to wot-discovery
issue 119 created in wot-discovery
McCool: Any objections merging it?
no objections
Christian: Can we mark it as work in progress?
McCool: You can raise additional points, if you want to. Or add editorial notes.
Farshid: There is an issue on github we can link to.
McCool adds comment about the issue being a WIP.
<cperey> sorry... not much to contribute to today's discussion so signing off until you return to the AR or geospatial use cases
McCool merged the PR
Geolocation
McCool: No additional points. Working on it.
<kaz> [ 5min break and then continue the discussion for today ]
McCool organizes some of the issues and prioritizes them.
McCool: Let's put all the terminology stuff into this one issue. (#119 in wot-discovery)
<kaz> Issue 119 - Revisit Partial TD terminology
McCool: Working on a PR. Fragment is inconsistent. So, proposed to change it.
McCool: Proposed to change it to TD element instead
McCool: Other terms to worry about?
No responses
#1042 error responses
<kaz> wot-thing-description PR 1042 - WIP: Add additionalResponses to Form
McCool: Right now we have in forms content-type. Expected to be input and output the same. Trouble is: Only the case for properties. You can also can different response from input.
McCool: ExpectedResponse only has contentType. No additional fields. This lead to problems in certain cases.
McCool: I think we need additional fields.
Christian: Do we have any examples for this?
McCool explains some cases where this my happen.
McCool: Thinking about updating my PR for it.
<FarshidT> The original issue has few examples: https://
McCool: Just wanted to point this out. We should discuss this.
Pagination
<kaz> Issue 16 - Handle huge set of Thing Descriptions (pagination, streaming, etc.)
<FarshidT> Example pagination: https://
Farshid: Different ways to do it.
Farshid explains the example
christian adds some suggestions to improve it.
Andrea adds a potential solution for it.
Farshid: Not very efficient way to do it.
Andrea: Diagram you showed before. Makes sense to have it as an attribute.
Andrea: Maybe it can hinder some developer.
McCool: What about other formats? For example CoAP? It may makes problems with other protocols.
Farshid: Do we really need this information?
Speaking of total
Christian: Anyone looked into the github api for it? Maybe it is an inspiration for this issue.
McCool: Do we really care about the count?
Christian: It is relevant for linked TDs.
McCool: If I get more than 100 back my query isn't specific enough.
McCool: In many cases I only care, if it is more than 1.
Andrea: sparql has its own pagination. REST doesn't have it, because it returns all the TDs and it may be a lot of TDs.
Discussion about using link-listed for it.
Andreas has agrees with christian that this may be a solution for certain protocols.
certain*
Christian: Farshid, do you need to skip pages?
Farshid: Not really.
McCool: Maybe we don't have a one fits all solution for it here.
McCool: Maybe we have different APIs for different use-cases here.
Farshid: Maybe we don't need the parent here.
McCool: Wrapper object complicates it.
Farshid: I don't really follow.
McCool explains details about it. With linked-list structure.
Farshid explains the specific use-case they have to christian.
McCool: Returning an array might be too big.
Farshid: We have pagination for it.
Christian: Why should we restrict ourselves to one TD?
Farshid: We need to do a lot of querying for it then.
<kaz> W3C API Overview
Kaz: thought I had already given this comment for this topic several months ago, there is a similar API named "W3C API" to access the W3C DB. With the parameters for (1) how many items should be listed on a page and (2) which page should be displayed. REST API. My point is: we can look at the document to get some insight. If we want to have a count, we also need a field for how much data should be included in one page.
McCool: We researched this already. We can compress it down. So, I would say: 16k maximum size of a TD.
McCool: anyways, that would limit our transaction size to 16k. What is the reasonable size for a transaction?
McCool: Let the implementation take care of it. In cloud may fine, but not on constrained devices.
Andrea adds details to the discussion with the discovery to it. Farshid comments some problematic use-cases to it.
McCool: We can take a look into. I tried to capture some points on it. We need to do an analysis on it.
McCool: I think we need some examples for it. How do we organize it? Complex discussion. How do we do a decision on it?
McCool: Feel like we have a lot of options with a trade-offs. So, we should collect it.
McCool adds some comments to the issue.
McCool explains some use-cases with different approaches to it.
McCool: What are the use-cases we need to research?
Andrea responds with the light-switch example.
McCool: Pagination doesn't apply to search results?
Farshid: no
Some discussion between andrea and Cristiano about sparql, xpath and jsonpath in context of wot.
McCool: If it just about the REST API, downloading the entire database is a rare use-cse.
Farshid: I disagree. You are putting more pressure on the server with a linked-list.
Farshid: In that case I can solve it with JSONpath.
Farshid: Maybe querying all on once may be not so useful.
McCool: last 5 minutes. Let's discuss next steps.
McCool adds comments to the issue #16
Farshid: At the current stage pagination is simple to provide.
McCool: JSONPath etc. have solutions for this, so we have to make sure not to duplicate it for these protocols.
McCool: We are out of time. We need concrete examples. Need PRs for it, as a starting point.
Farshid: Agree, I can write it down.
McCool: I keep to open it for other people.
<kaz> McCool's comments for issue 16
<kaz> [adjourned]