Spatial Data on the Web IG F2F - Day 2/2

20 February 2018

Meeting Minutes

SSN issues

Pull request https://‌github.com/‌w3c/‌sdw/‌pull/‌982 fixing typos and wrong links

Armin: I was wondering whether I can accept the pull request right away. What's the process. Do we need to update the Note?

Francois: [explains the process, difficult to update the Recommendation, errata document, possibility to update the Editor's Draft in whatever way we want]

SimonCox: There's nothing that stops us from making any change in the Editor's Draft. The issue is that nothing that we do in the Editor's Draft would stand any chance of going into the Recommendation

Francois: Right, the IG is not chartered to make normative changes to the document.
... The Editor's Draft is supposedly owned by the IG

Armin: OK, we'll get back to that. So I can approve the pull request here.

Jeremy: No objection.

Armin: OK, merged.

Pull request https://‌github.com/‌w3c/‌sdw/‌pull/‌1000 in response to https://‌github.com/‌w3c/‌sdw/‌issues/‌998 and https://‌github.com/‌w3c/‌sdw/‌issues/‌999

Armin: Hard to follow what the actual changes are, because the serialization changed. What's the proposed changes in SOSA/SSN?

mlefranc: Actually, I just added the RDF/XML serializations so that we keep track of them from now on.
... The pull request implements a fix in SSN DUL
... Generation of the files done with Protégé
... Issue #998 is harder to fix.
... If you're loading the SSNX ontology in a reasoner, you wouldn't be able to add an individual to those classes. Issue explains it all.
... I just relaxed axioms (1) and (2).
... sosa:hasValue now subProperty of oldssn:hasValue, same for sosa:hasResult
... The only thing that is implemented in the pull request is to track the changes in the errata.
... Third issue is that there is a disjunction between ssn:Property and sosa:FeatureOfInterest
... I think the reason why there is an unsatisfiability is that we need to relax a couple of other axioms.
... I remember doing it back in April last year, but things changed afterwards and that fell through the cracks.

<Zakim> jtandy, you wanted to confirm that the “complete” files are just adding an RDF/XML encoding of existing material?

jtandy: The complete files added in there are just RDF/XML encodings of stuff that were already in there?

mlefranc: This is the equivalent to the Turtle document.
... No base URI in the versions online, this new version keeps the base URI.

jtandy: You talk about relaxing axioms. Is that backward compatible?
... If a document conformed with the previous set of axioms, does it still conform to the new one?

<SimonCox_> 'conform' has no real meaning in RDF ...

<arminhaller> not inconsistent

mlefranc: Yes.
... A valid ontology with the previous document is still a valid ontology with the new one.

arminhaller: My understanding is that we need to list the changes in the errata document.

<tidoust> Errata document

Francois: Question is whether we're talking about normative changes, or fixes to the spec. We'll need to add an entry to the errata document. The fix itself can be incorporated in the Editor's Draft.

mlefranc: It affects a non-normative section.

jtandy: So, that's all good. That falls within the scope of the IG.
... We cannot update the published Recommendation, but it can go straight in the errata, and we can update the Editor's Draft

mlefranc: A confirmation email would be good

brinkwoman: What about ontology files? Can they be updated?

Francois: Yes, no problem. My responsibility is to keep them in sync with what the spec says, but there are no official process we need to follow, we can update these files whenever we want.

<tidoust> [Some discussion on where the errata document lies. Armin and Jeremy would prefer the document to appear in the GitHub repository. Francois will move it there]

Proposal https://‌github.com/‌w3c/‌sdw/‌issues/‌1004 to include hasUltimateFeatureOfInterest for the property-chain ( sosa:hasFeatureOfInterest o sosa:isSampleOf )

Action: François to migrate the errata document to the GitHub repository

<trackbot> Created ACTION-383 - Migrate the errata document to the github repository [on François Daoust - due 2018-02-27].

<SimonCox_> This diagram summarizes https://‌github.com/‌w3c/‌sdw/‌issues/‌1004#issuecomment-363990213

jtandy: Great, then we can integrate the change to the errata document in the pull request and merge

SimonCox: The proposal is to make a substantive change to add an ultimateFeatureOfInterest.
... Clearly, this is beyond the scope of the Interest Group.
... However, I'm bringing this now. Is it possible for the IG to publish new axioms in a different namespace, as a W3C Note?
... What kind of artifacts can the IG publish?

Francois: The Charter envisions the possibility to create a separate Note, so that's perfectly ok.

SimonCox: Wondering if we can reuse the same namespace, or whether we need another one. Making changes in the SOSA namespace may not be possible.

mlefranc: From what I can tell, I don't see any problem with adding a new document to the SOSA/SSN namespace.
... Maybe an external module defined in a separate document.

jtandy: In terms of the namespace, so long as it doesn't change how the rest of the spec works, I believe it's just a slightly bigger version of what we've just been talking about.
... People can ignore this extension.

SimonCox: It does not disrupt anything, right.

jtandy: The complementary document that would describe this extension would be published on the side.

arminhaller: We are allowed to create a new Note that defines hasUltimateFeatureOfInterest. And we could create an ontology file that imports the SSN ontology and adds the property. Is that correct?

<SimonCox_> That is what I heard
... That new file would use the same namespace.

jtandy: That's my understanding.

SimonCox: Moving onto part 2 of my question, take a look at my diagram, and let me know what you think

arminhaller: I suggest to park it for now, and discuss that now that we know what we can do from a process perspective.

RaulGarciaCastro: To increase the adoption of the SSN ontology, publishing a separate document is confusing. I would collect issues and possible solutions, but would not publish other non-normative documents
... For us, it might be clear, but for other people it would be very confusing.
... When we have a number of issues, we can start writing a Note.

SimonCox: Now we're getting on tactics and minuting. This proposal was prompted by feedback from people using the ontology.
... Someone contacted me saying she found it hard to use the SSN ontology because it was missing that feature. Hence the proposal to solve this. If there are barriers to adoption that can be solved by amendments like this, we should remove these barriers. People in the community would more easily adopt the ontology with these changes.

jtandy: It seems better to add extensions in a controlled namespace than having everyone publish extensions everywhere.

SimonCox: Right.

RaulGarciaCastro: We can propose solutions as an Interest Group. We shouldn't create a new version. Unless we plan to publish a new Note every year, but that's confusing.

arminhaller: We know the process now. We may need to come to a resolution on what we're going to do. How we will implement this Note. What I hear is that we want the Note, but the question remains as to whether we also provide the implementation in terms of an ontology file.
... Next issue also somehow relates to a process issue

Proposal https://‌github.com/‌w3c/‌sdw/‌issues/‌1006 to move hasProperty from SSN to SOSA namespace

arminhaller: Just wanted to quickly raise that issue. We discussed that issue for quite a long time initially.
... In the very beginning, the decision was made that we keep specific properties but that generic properties would be confusing.
... The difference here is that this would be a change to the normative part, which is not possible.
... Assuming we would agree on this, what this would mean is that the term would move from SSN to SOSA. It could make it into the Note. We could create an ontology file that imports the SOSA ontology.
... And defines the term. The SSN ontology would deprecate that term.
... That's a different issue, because it touches on normative parts of the document.
... That could be done in a companion note

<SimonCox_> Would you also say sosa:hasProperty owl:equivalentProperty ssn:hasProperty . ?

mlefranc: I would be in favour of what has been proposed. Ironically, we would end up with two documents that end up defining terms in the same SOSA namespace.

<Zakim> tidoust, you wanted to mention the possibility to re-charter a WG

<SimonCox_> mlefranc identifies a technical packaging/publishing issue but I'm relaxed about that

Francois: It's always possible to re-create a WG to publish a second version of the SSN Recommendation. That would make things cleaner from a normative perspective. Or the document could be put in scope of an existing WG if we can find one.

jtandy: So idea here is to move a term defined in the complex namespace to the simple namespace.

arminhaller: Yes, but we don't have a resolution here. We can continue the discussion on GitHub. If we end up with a resolution on GitHub, we can think of creating a separate Note, then a separate ontology file, and then decide whether we can work on a new version of the SSN spec.

jtandy: OK, so continue the discussions. Collect issues and propose solutions in a Note, and then consider whether this warrants the chartering of a WG

SimonCox: I'm concerned we're proceeding over Raul's objection

RaulGarciaCastro: I'm fine so far

jtandy: Armin, do you want a resolution on how we proceed?

arminhaller: Continue disussion. Find resolutions on GitHub. Collect solutions in a separate Note. May work on separate ontology files. Eventually may lead to the chartering of a new SSN WG if there's enough material.

PROPOSED: Re. SSN issues, idea is to: continue disussion, find resolutions on GitHub, collect solutions in a separate Note. Solutions may be implemented in a separate ontology files. Eventually this may lead to the chartering of a new SSN WG if there's enough material.

<jtandy> +1

<RaulGarciaCastro> +1

<brinkwoman> +1

<SimonCox_> +1

<arminhaller> +1

<tidoust> +1

<mlefranc> +1

<billroberts> +1

<RobSmith> +1


<MichaelGordon> +1

Resolved: Re. SSN issues, idea is to: continue disussion, find resolutions on GitHub, collect solutions in a separate Note. Solutions may be implemented in a separate ontology files. Eventually this may lead to the chartering of a new SSN WG if there's enough material.

jtandy: The knock-on from that is that you will work on finding a consensus on the two issues that you have on GitHub.

arminhaller: Yes.
... On a different front, the outstanding issue around starting our SSN primer, haven't started yet. Busy time of the year. I will kick things off next week hopefully.
... Happy that we've made progress on the pressing issues on SSN.

<RaulGarciaCastro> jtandy Which day do you have the slot?

jtandy: Thank you Armin. I would note that we have a small slot on the plenary of the OGC Technical Committee in Orleans next week. If you want Linda to raise the SSN Primer there, just tell us, and provide some material.

<mlefranc> thanks, bye !

<SimonCox_> bye

<RaulGarciaCastro> Bye

<ChrisLittle> bye

<arminhaller> thanks, bye!

Stats on the Web BP

See <billroberts> agenda for Stats on the Web BP

billroberts: Want to clarify what we're aiming to do, and what we're not, especially because it's potentially a very large subject.
... I'd like to gather existing work that we should consider.
... And then get on some consistent way of doing things.
... Obviously, we're working within the Spatial Data on the Web IG, so we should pay particular attention on the intersection between statistical data and spatial data.
... It's "on the Web" as well.
... Also, what do we mean by statistics. Government data, but also scientific statistics, such as meteorological data.
... Are there other types of data that we should take into account?
... I think it should be narrower than "all data", but most data has some statistical component.

ChrisLittle: We don't want to reinvent wheels. What is distinctive for statistical data compared to regular data on the Web? That's the key thing

<billroberts> Evangelos - looks like our phone connection has dropped - just dialling in again

<billroberts> we are back but looks like you have gone - hopefully you can rejoin soon

ChrisLittle: One of the ways of finding the distinction is by looking at the existing best practices to identify requirements that are clearly specific to statistical data.

MichaelGordon: Do we feel that we have the mandate to look at statistical data in general, or at statistical data in relation with spatial data?

ChrisLittle: I think that restricting to geospatial data, we might be missing something. So actually, I think we should look broadly initially.
... That's the reason why I raised a use case about time.

billroberts: Also, I have the feeling that it does not narrow it that much.
... Most characteristics should be similar. Geospatial data actually extend the scope.

MichaelGordon: I was more thinking of narrowing down the range. Also lack of authority among participants perhaps

ChrisLittle: If you look at the main difference between data on the web and spatial data on the web best practices, that's the CRS.

<billroberts> https://‌lists.w3.org/‌Archives/‌Public/‌public-sdwig/‌2018Jan/‌0048.html

billroberts: Andrea went through the list of data on the Web best practices, see email.
... A structure for a best practices document could be to follow the Data on the Web Best Practices document, and extend each section with "if you're dealing with statistical data..."

jtandy: Initially, we thought we'd do that for Spatial Data on the Web, but then decided against it.
... In some cases, nothing to add. Also, it started with metadata and we did not want to emphasize that first in the Spatial Data on the Web Best Practices
... For sure, you should go through the best practices and check whether we want to add something or not.
... All data on the Web Best Practices apply.
... You may want to add extra things.

brinkwoman: Did Andrea find anything missing yet?

billroberts: It's just a starting point for discussion right now.
... The "on the Web" bit is also where we can come up with advice.
... Current practice in the UK for government data is that figures will come up every year, published as a PDF document accompanied by a couple of spreadsheets. That's how it gets done right now.
... Lack of machine readable access to it. Lack of interoperability between different publishers.
... That's an example of an area where we could point out best practices.
... In the scientific sector, it could be large files, probably more machine-readable, but subsetting it is difficult.

ChrisLittle: That's one aspect. Also about understanding what the data is.
... For instance, wind speeds, you may use them, but not realize that it's a mean wind speed, and that there are extremes as well that the data do not convey.

billroberts: Adding URIs to this data allows you to associate definitions.

ChrisLittle: Standard controls and vocabularies. Don't invent your own definition of arithmetic mean.

billroberts: Also, if you find multiple properties that match what you wnat to expose, how to choose?

ChrisLittle: The authority might be important. The authority might exist but data has not been updated in 20 years

jtandy: When we started the work on SDW BP, we thought we'd have a lot to say about ontologies, but then there was not to say in the end. Just "look at your community" in essence.
... Also, the CSV on the Web work. One of the key motivation for that is people publishing statistical data in spreadsheets.
... Quite a lot of material in CSV on the Web demonstrating statistical data.

<Zakim> jtandy, you wanted to mention CSVW

billroberts: I mentioned in the agenda examples of works we're aware of.
... It's not that we need to produce a comprehensive list.

<jtandy> CSV on the Web Primer: https://‌www.w3.org/‌TR/‌tabular-data-primer/
... The way that we could take those into account.
... One problem is we haven't got any proper statistician in this group. We want to make sure that our output gets reviewed by experts.
... I've been getting in touch with a few people, UK stats.
... Linda managed to get someone from the Dutch bureau of statistics.

ChrisLittle: Also look at ICSU (possibly CoData)

billroberts: Best to take actions to engage people. Happy to check with statistical government agencies.
... SDMX people as well, been discussing with someone.

RobSmith: Would it make sense to contact Geovation as well?
... Looking at WiFi and buses, presumably they have statisticians looking at data

MichaelGordon: The Geovation crowd themselves probably does not do that directly.

billroberts: Good group to engage through other UK based communities

jtandy: I remember that Kerry Taylor did some presentation about stats in Australia

billroberts: Yes, used to work with the Australian Bureau of Statistics. I believe Jo works with them now

billroberts: Looking at our work program now

<billroberts> https://‌github.com/‌w3c/‌sdw/‌blob/‌gh-pages/‌stats-bp/‌work-plan.md
... I proposed the work plan. No objection.
... By the end of this month, we should aim to have a first version of the use cases document.
... That's not entirely unfeasible. I don't think that will be a closed document.
... Then I was proposing to turn these use cases into requirements by the end of March.
... Merge the two by end of April, and then draft best practices.

jtandy: I think we spent too much time on use cases before doing the spatial data on the web best practices
... I would not try to finish use cases upfront. I would do things in parallel.
... You lose nothing going through the Data on the Web Best Practices.
... It will stimulate you in finding use cases, actually.
... Is statistical data special for each one? That's a good filter.

ChrisLittle: Right. We should simply add a couple of use cases to the document to have something slightly more complete in terms of coverage.

jtandy: First Public Working Draft really just shows that you've started working, doesn't have to be finished or polished.
... The more finished it looks, the fewer people will come and help you.

ChrisLittle: Putting controversial things in there can actually help getting feedback

jtandy: A regular release schedule proved useful. An iterative approach.
... It helps focus the minds of people involved in the work.

brinkwoman: It would be nice if you have a list of best practices that you want to write in the right order beforehands.

Francois: I was about to suggest the opposite in a way. Don't worry about the order to start with. Just keep in mind that the order will change, so make it easy to do so and don't spend time on intro text.

billroberts: OK, I will revise that draft document accordingly.

<tidoust> [side discussion on the order of best practices. You have to see the draft list to have a feeling of what will be possible]

jtandy: The current order in the SDW best practices can be applicable to statistical data. You may have extra things to look at though.
... Some things we wrote in Web principle best practices were written because they were missing from the Data on the Web Best Practices, which could not easily be updated.

Discussion in 2017: https://‌www.w3.org/‌2015/‌spatial/‌wiki/‌BP_2017_reordering_proposal
... Question as to whether you reproduce them or reference them

jtandy: Looking at the first best practice, we were willing to have people think in terms of giving an identifier to a real postbox.
... Not sure that applies to statistical data.

billroberts: Statistical data will have a reference to a spatial thing most of the time.

ChrisLittle: You also have the concept of a sample, which is a subset of your population.

jtandy: All of our dimensions should be uniquely identified in other words

billroberts: Yes.

jtandy: So going through the existing best practices is useful to identify what to say for statistical data

brinkwoman: Is it possible that additional best practices for Spatial Data on the Web will pop up during discussions?

ChrisLittle: possibly

RobSmith: You touched on modeling. Simulated data vs. actual data.

billroberts: A big area that Chris is highlighting. What you are measuring.
... SDMX has the mechanism but people do not use it a lot. What your statistical measure is.
... Is a mean, median, percentile? What was the sampling approach? What was the population you were sampling from?
... Lots of surveys. How were they designed? How did you choose the people you asked?
... How did you correct the values to take the rest of the population into account?
... With a URI, you can link back to a methodology document.

jtandy: That's a good start.

ClemensPortele: Wondering about the liaison with the Semantic Statistics Community Group. The CG seems dead

Francois: Evangelos represents that group here.

billroberts: Essentially, the group has done nothing. But people in the group are active in the field.

kalampokis: All the people who are currently in the group are working on publishing and reusing statistical data using linked data technologies.
... They say that they are interested in having discussions.
... I will send an email to coordinate. Whether we want to have a parallel discussion in the CG or have the discussion in the IG directly is up for discussion.
... I can help with coordination.

billroberts: Maybe a good way to do it is to ask them a specific question. Once we have a draft document, "what do you think of this?" rather than "please engage".

kalampokis: I agree

billroberts: Any other comment?

jtandy: Is the motivation to check an existing body of work and encourage people to share the same approach, or are you trying to identify a wider pool of best practices that you believe should apply?

billroberts: My objective is to get different publisher doing it in a more compatible way.
... Bringing good expertise is one aspect of it.
... The reason I care about interoperability is that the data would have more value, which in turn would mean that people would have more incentive to publish data.

jtandy: So, you have working practice and you hope to identify other working practices and find some harmony among these.

billroberts: To what extent do we document practices followed vs. recommendations for new practices?

Francois: That question always comes up in best practices groups. No real answer. "If there's no solution, that means there's no problem" (copyright Shadoks)

jtandy: The CG must be able to identify examples that they're working on, and other sources.
... That makes me think of an interesting question: the statistician will be the ones working with the data published, not the ones publishing data.

ChrisLittle: I think it's too simple to say that the document is targeting publishers only. Need to take a user perspective.

jtandy: Looking back at Spatial Data on the Web Best Practices, we decided not to write best practices on how you might consume, or reuse your data.

billroberts: Statisticians are often responsible for the data that gets published

Francois: One of the best practices could actually be to publish raw data along with the analyzed data so that other analyses can be made on the data.

<tidoust> [discussion on anonymizing data]

MichaelGordon: Does that bring some requirement about reversal of data? If you're publishing statistical data, choosing appropriate anonymisation techniques (considering the GDPR)

<jtandy> (We talked about being able to trace from an aggregate dataset to the raw dataset, e.g. using PROV, where the raw data )

billroberts: That's a good question. I'd like to refer to things where that's been considered.
... It's a big deal in government data: the methodology you used to anonymize the data.

<jtandy> (Cont... raw data may not be published online, e.g. because it is not anonymous / personal data)

RobSmith: I looked at privacy issue when we published a personal tracker on the phone. I haven't thought of revealing something else. Example of the military bases with joggers data.

jtandy: In terms of what you want people to do with analytical data. Are you expecting people to do statistical analyses online or to take their own copy and work offline?

billroberts: At the moment, it's more an offline model.
... There is a question of easy access to the latest data, and the ability to make subsets.
... One of the best practices we had for spatial data was about supporting different formats to ease reuse. This will apply here as well.
... You'll want to facilitate downloads in different formats.

jtandy: Don't assume that one format is enough was one of our assumptions. Also applies here.

billroberts: Yes.
... Sometimes, people need a clear definition of what the terms are, sometimes, they don't care. Being able to ignore that info when not needed and find that info when needed is useful.

Francois: One side point that publishing the Use cases & requirements document as a First Public Working Draft is not required per se. Can remain a document on GitHub, or you can fold it into the Best Practices document to start with.

jtandy: That's fine if you can cross-reference the use cases with a permanent identifier in the end

Francois: Right, what I mean is that you don't have to follow a specific order. That can be done later on, as needed. Still a good practice to publish a FPWD before publishing a Note, even though the process allows to publish a Note directly.

<billroberts> https://‌github.com/‌w3c/‌sdw/‌blob/‌gh-pages/‌stats-bp/‌draft-use-case-list.md

billroberts: OK. Final thing I wanted to talk is our list of use cases
... It would be interesting to get suggestions about use cases that we don't have. Also about styles.

ChrisLittle: And whether we need a formal structure.
... Certainly my use cases, I can turn them into requirements, and write a use case scenario.

billroberts: Seems useful to provide a bit of context to people who are not experts

MichaelGordon: It provides the "why" it is important.

Francois: The why will appear in the Best Practices. Need to strike a right balance between spending time on use cases and on the best practices. Should probably be skewed towards the best practices. Iterative process.

billroberts: Rough summary of what's in here so far.

<billroberts> Evangelos: we are presenting and talking over https://‌github.com/‌w3c/‌sdw/‌blob/‌gh-pages/‌stats-bp/‌draft-use-case-list.md

ChrisLittle: First one is about representing statistical parameters.
... I see people still inventing conventions for describing parameters

billroberts: Always about the definition of what you're measuring.

ChrisLittle: Sometimes, it's hidden in the unit, other times in the name, other times in an extra attribute.
... In fact, 25 years ago, we mixed all 3.

<jtandy> The statistical measures that Chris mentions are published as linked data here: http://‌codes.wmo.int/‌grib2/‌codeflag/‌4.10

billroberts: Right, it should be clear for everyone what a particular number represents.
... GRIB2 Code Table 4.10

ChrisLittle: This was the problem we had, and this is the approach we took. Requirement is along the lines of a standard mechanism of annotating.

jtandy: One of the things you wanted to do was to reference vocabularies.
... One of the challenges is to relate the notion of average in different domains.

ChrisLittle: Next use case is around representing temporal data.
... Aggregating values per month for instance, where months don't have the same number of days.

<tidoust> jtandy: Daily minimum probably goes from 8:00am to 8:00am for instance.

ChrisLittle: Exactly, and the minimum could be before midnight, which might upset people.

<tidoust> [discussion on possible mechanisms]

billroberts: Next use cases on area profile. Some kind of collection of data for a school to decide whether to send your child to it or not.
... A lot of it is around good identifiers for places and discovery of data, trust of sources.

MichaelGordon: example of policy makers. Simple tool made to compare e.g. employment rate in different cities. And then refinements.

billroberts: A lot of local authorities spend a lot of time doing this.
... Can we understand the data enough to compare it?
... Finding places by criteria
... Next one is about slicing and dicing a statistical data cube.
... Very commonly, you want to arrange a subset including possible values on a given dimension.

<jtandy> https://‌github.com/‌w3c/‌csvw/‌blob/‌gh-pages/‌examples/‌rdf-data-cube-example.md

jtandy: The RDF data cube example I wrote for CSV on the Web could perhaps be relevant here.
... It shows how we constructed the metadata that goes with the CSV to create an RDF data cube.
... This particular dataset was human processed. Other stuff happen upstream to this.

billroberts: Some use cases about sharing identifiers. Consistent identifiers. Somebody has to manage these codes.

jtandy: Is that not a generic practice about publishing data?

ClemensPortele: That's what I was thinking too.

billroberts: Maybe nothing specific to statistical data.
... Usually done well for geographical areas, and badly for everything else. Things would be a lot more interoperable if people did that better.

ClemensPortele: When I look at the area profile use case, you need to bring things together which supposes common vocabularies.

jtandy: The use case is to use data because they understand how to aggregate it.

billroberts: It's not a technically complex thing, but social issues usually stand in the way.

<tidoust> [looking at "registers" in the UK]

billroberts: Also use case for publishers. When you're preparing the data, you need to be able to identify what the right identifiers are for your data. Maybe same use case
... Another one that has come up in various forms is provision estimates. Statistical organizations usually improve on published data after an initial publication. Example of GDP data.
... 3 estimates of GDP in the UK for instance.
... You might end up with 3 different values for the same thing.

ChrisLittle: Same concept in meteorological where we label data with Level 0, Level 1, and Level 2.

MichaelGordon: Requirement is about being able to version a statistical set of data that you may be publishing

billroberts: Yes.

jtandy: Already covered by existing best practices

<billroberts> Evangelos - just re-starting webex here

Francois: In the end, instead of new best practices, you might end up with techniques to apply best practices to statistical data. That would be a fine document too.

billroberts: Another use case about comparing data across countries.

jtandy: In the CSV on the Web work, that is one of the use cases that motivated Jeni's work.

billroberts: Comparison of geographical data. Specific case of an existing best practice
... The next one is too broad, because it encompasses all others. Represent statistical quantities in RDF.
... Here are examples of data we have, how to represent them?
... OK, that's kind of where we are at the moment.

Francois: Previous discussion on data anonymization and publication of raw dataset could be turned into a use case too.

ChrisLittle: People have some data and they want to annotate it in some way to identify the uncertainty. That would be a valid use case as well.
... Both quantitatively and qualitatively

<tidoust> [example of temperature measurements excluded when computing averages for climate change computations, e.g. due to someone mowing the lawn nearby]

billroberts: Probably some data on the Web best practice that talks about it.

<jtandy> ref the project that Jon Blower was involved in about annotation metadata ... the CHARMe project: http://‌charme.org.uk

billroberts: So I've pretty much reached the end of what I wanted to touch upon.

RobSmith: One comment about use cases. I try to think about scenarios that are realistic. What I find is that some of the details can easily be lost. What I found useful to me was to have a list of benefits linked to each use case.
... The use case may have more than one benefit.

<Zakim> jtandy, you wanted to ask if the BP template will be reused?

jtandy: Will you use the BP template?

billroberts: General plan is to follow the same plan, essentially yes. Any lessons learned?

jtandy: I think that would help create a generic feeling across the best practices that we write.
... Also we reused the benefits from the Data on the Web Best practices document

Back to the Spatial Data on the Web Best Practices

brinkwoman: In one of the calls, I mentioned that it could be a good idea to create a tool to do validation or conformance testing
... We usually do it at Geonovum.
... Possibly it's not easy or possible to create an automated checker.

ClemensPortele: My view is that, at this level, it is not possible.
... Best practices require some interpretation.
... You can only do that with specifications.
... What might be possible to do is to have a checklist of things that you could check.

Francois: Agree with starting with techniques, a checklist. An automated checked is hard to achieve and time consuming, and often time boxed (techniques evolve). Example of the mobileOK Checker.

<tidoust> [also question of hosting the service]

<tidoust> [discussing which best practices are being addressed in WFS]

brinkwoman: It seems it's useful to create a checklist.

ClemensPortele: Another direction would be to make explicit the best practices you follow, but I don't think we're at that stage.

RobSmith: [example of GPX validation when crossing the Greenwich meridian, with 1e-04 numbers making the value invalid]. Validation is useful.

<tidoust> [discussing possible validation tests]

jtandy: (goes through the agenda items for the afternoon)

<jtandy> https://‌www.w3.org/‌2017/‌sdwig/‌meetings/‌f2f-1.html

jtandy: The goal for the afternoon is to identify the input to the funnel (once we have the green light to add to it)

Video Geotagging format for Electronic Maps

<jtandy> https://‌www.w3.org/‌2018/‌02/‌19-sdw-minutes.html#ResolutionSummary

jtandy: (reads the five resolutions)
… only work that goes to a standards track will end up in the funnel
… we have a process in place

<tidoust> Evaluation phase in the strategy funnel

tidoust: this is the process to decide whether something is ready for chartering

<tidoust> Rec track readiness best practices

tidoust: provides a check list - is there a clear problem statement? explicit success criteria? do we have the right participants and is there an ecosystem? etc.

jtandy: That is the re-cap I wanted to go through. Rob, is it clear enough to you as the first candidate?

RobSmith: Yes, it answers my questions. I don't want to waste peoples time.
… The current proposal is at an early stage, currently about 6 weeks of work from one person. Will it work in a browser?
… Funding is important, too. Funding application pending. I would be interested, to understand early, if there is interest in the group.

tidoust: This goes back to the question of ecosystem and participants.

RobSmith: I gave a short presentation at the last plenary. Would it be worth repeating?

jtandy: Yes, please.



RobSmith: Slides are brief and contain only the main points (10 minute presentation)

RobSmith: New slide set has additional detail (12 slides).
… Work was triggered by a call from OSGB to show the location of people in a tv show. Geotagging in pictures is available, but not really in videos.
… Show particular events in the video on a map. Not just the location, but also other presentational input (e.g. direction).
… Also the capability for annotations.

MichaelGordon: Location and direction are essential also essential for AR

RobSmith: I set up a tech demo for this. Came across the VTT mechanism for subtitles on You Tube. Based on a standard.
… Text-to-speach support. There is an existing mechanism for annotations that could also be used as a framework to add geospatial information.
… Did a comparison between photo and video. Standard mechanism for photos (exif, many cameras support it out-of-the-box), but not for video (except add-on for high-end equipment)
… But could we use the smartphone (can capture video and has sensors for spatial information)

<PeterRushforth> hi don't have access to webex info, sorry

tidoust: This may be a reason for not having the information in a separate file. Device manufacturers will likely prefer an integrated approach.

RobSmith: I think GoPro records location information in a separate text file. But I have not checked in detail.

<jtandy> https://‌www.w3.org/‌TR/‌web-packaging/

jtandy: Web-packaging may help here

tidoust: My point was to be careful how to present this.

RobSmith: An extra file helps to leave MPEG as it is

MichaelGordon: Does it require hardware manufacturers to support it?

RobSmith: Not necessarily, but it could help
… you can do it in software "today", and in the hardware devices in the future

jtandy: This is about spatial data on the Web, so this fits our mandate

RobSmith: The idea would not be to support just a single maps API (Google Maps, Open Layers, Leaflet), but support multiple ones

ChrisLittle: An abstract spec would help to clarify the concepts independent of the implementation and identify conformance classes

RobSmith: Use case examples - coastguard/mountain rescue, area survey

ChrisLittle: Have you considered temporal aspects? E.g. which areas have been covered?

RobSmith: Yes.

(discussion of work starting in the OGC UAX group)

tidoust: A key benefit of a standard is to share data, so this should be included in the use cases.

RobSmith: The third use case (swarm monitoring, operating drone within a swarm) does this

RobSmith: This is about enabling the use cases, not providing a (closed) solution

RobSmith: Francois asked the queston how a web app would use it.
… (shows <track> element in a <video> element and a sample modified VMT file)

tidoust: Interesting that you are focussing on rendering, not the data itself

RobSmith: The rendering is important, the line has a particular meaning (it is a multilayered problem and we do not have time to go into detail)

tidoust: Is there any relationship to the "describing moving objects" deliverable?

jtandy: No, we have no information yet on that deliverable...

RobSmith: (shows tech demo)

RobSmith: Any frame in the video is associated with a map - it is synchronised

jtandy: I think we all agree that fits, it could be for OGC and/or W3C, it requires a standard

PROPOSED: Add Video Geotagging Protocol For Electronic Maps to the funnel

ScottSimmons: There are probably some OGC members that have something that is similar, so we should engage them early.

jtandy: Patent free is also important.

RobSmith: No issues, I want this as a open standard to see adoption.

billroberts: The evaluation of existing standards would come as part of the evaluation process?

tidoust: Yes!

<MichaelGordon> +1

<tidoust> +1

<jtandy> +1

<brinkwoman> +1

<billroberts> +1

<ScottSimmons> +1

<RobSmith> +1

<ChrisLittle> +1

Resolved: Add Video Geotagging Protocol For Electronic Maps to the funnel

jtandy: This should be added to the first phase (exploration). Scott, how should Rob engage with OGC members?

ScottSimmons: Will do a call for information and reach out to some members this/next week

jtandy: Next plenary call we will include and update on the status. Rob can you track those actions?

RobSmith: Yes


jtandy: (welcomes Hugo Ledoux)

Hugo: (shows CityJSON slides)
… mostly following the OGC work, but not actively involved
… CityGML - for buildings and much more (basically everything inside and outside of cities)
… 5 levels of detail (LoD)
… adoption of CityGML not great, software support is limited, hard to work with CityGML files
… I immediately convert it to OBJ to work with it
… Data model is good, GML encoding hard
… Some time ago I looked into a JSON encoding for the data. Shared on the CityGML list, a mixed reception.

tidoust: Is license of CityGML an issue?

MichaelGordon: No, not with the OGC license

Hugo: Difficult to work for students to work with CityGML files, easier with CityJSON
… Now mapped all the CityGML modules
… Did it manually and not using an automated encoding rule. The reason is the complexity of the conceptual model.
… Some constraints (e.g., all geometries are in the same CRS)
… For now left out raster, DTM
… Added a metadata object
… Topology is preserved, vertices are referenced from the city object geometries
… Surfaces can have semantics
… Currently only links within the same file
… (discusses differences to GeoJSON)
… some similarities to TopoJSON
… a JSON schema is available
… We wrote software to support CityJSON - from the beginning
… Claus Nagel implemented CityGML <-> CityJSON in citygml4j library
… I do not plan to parse a CityGML file ever again, just convert it to CityJSON and use that
… Important to involve developers early in the process

jtandy: Questions for Hugo?

tidoust: I ask stupid questions as I am new to the map aspect, look at it from the Web perspective. Some discussion on the mailing list.
… What is the relationship with maps. How does it fit together?

ChrisLittle: You are a bit restrictive. Map is LoD 0.

tidoust: Map is not a 2D map for me.

Hugo: different cities maintain different feature types beside buildings (terrain, roads, etc).
… I prefer to talk about data, not maps, but the data can be used for maps etc.

ChrisLittle: Maps are typically the output of a computation using the data

ChrisLittle: Since you have offended a few people in OGC, would you work with them.

Hugo: Yes, and I like the model of CityGML. CityGML 3 will also look at other encodings.

billroberts: This reminds me about the CoverageJSON discussion we had earlier in the SDWWG.

<jtandy> CoverageJSON: https://‌covjson.org

billroberts: CoverageJSON never got that far due to open questions with regard to "webiness" - can you link into it, can you break it down into manageable pieces

Hugo: There is work with using CityJSON in WFS. Currently CityJSON is not tailored for the Web, but it may evolve to address this.

jtandy: When you look at CityJSON, does it seem like an encoding that could be supported by WFS 3.0?

ClemensPortele: I don't see that as being mandated by a conformance class today, but you could write an extension.

jtandy: So it would fit in the OGC ecosystem?

ClemensPortele: Yes!

<billroberts> CoverageJSON: what I meant to say is that in the SDW Working Group we didn't have time to investigate in detail about identifiers and API mechanism for easily retrieving subsets. For the core use case of CoverageJSON, to have a Javascript friendly (and hence web browser friendly) representation of coverages, then I think CoverageJSON achieved its goals very well

jtandy: Is it available under an open license? Patents?

Hugo: It could be made available under an open license. No patents.

jtandy: Why should it be a standard?

Hugo: Not interested in fighting for it, but a label as a standard would help to see adoption by government organisations.

jtandy: Any others beside Claus Nagel and you that would work on this?

Hugo: Not at the moment.

jtandy: So some community building is needed, not yet ready for chartering. Further down the line than the video geotagging.

jtandy: It would probably fit with OGC due to the closeness with CityGML. But visibility for the Web community may be relevant, too.

tidoust: Yes, closer integration with Web Browers would indicate that it would fit in W3C, too, but right now this does not sound like the main focus.

Hugo: At least for now. Some concerns that the CityGML development is not developer focussed, W3C seems to be more developer driven.

MichaelGordon: One thing that should be discussed in an incubation phase are the SDW best practices.

Hugo: Yes, I welcome input on this.

tidoust: This could also be an indicator for a joint process.

PROPOSED: Add CityJSON to the funnel in the incubation phase

<jtandy> +1

<brinkwoman> +1

<tidoust> +1

<RobSmith> +1

<billroberts> +1


<ChrisLittle> +1 on behalf of Hugo

<MichaelGordon> +1

Resolved: Add CityJSON to the funnel in the incubation phase

jtandy: What do you need for the next phase?

Hugo: Feedback from a Web perspective. How to deal with large files. All discussions should occur on the GitHub repository.

(agreement to continue using the GitHub repository)

jtandy: Are the issues registered as issues on GitHub?

Hugo: Not yet, but I will add them.

brinkwoman: Should Rob and Hugo join the IG as an invited expert?

tidoust: It would be good. I will work with them on this.


jtandy: We want to find out how and where MapML fits.

PeterRushforth: The Maps for HTML CG started as a result of the W3C/OGC workshop in London in 2014.
… Wouldn't it be good to use as an HTML author just a regular HTML element to put a map on a web page?
… In 2014 I had a mandate to work on this as looking for support to the idea. We made some progress since then, including developer resources in a GitHub organisation.
… Map Markup Language is the key specification I am going to talk about today.

<brinkwoman> https://‌maps4html.github.io/‌MapML/‌spec/

PeterRushforth: MapML is a "DOMable" format.
… Oriented towards browser-developers, not web-developers.
… Extend the HTML to support maps in the traditional web mapping sense. Building on top of the "image map" concept from HTML, i.e. a more general <map> element.

<ScottSimmons> Scott is back - sorry, much worse roads than I had hoped

PeterRushforth: The draft makes use of Web Components and specifies a customized <map> element. A draft implementation exists.
… (shows JavaScript implementation demo)
… NRCan uses MapML, mainly in demos
… (demonstrates adding an additional layer to a map)
… In OGC Testbed 13 both conceptual work on MapML (Engineering Report) and another implementation.
… Could bring maps closer to the W3C community.
… Less interest in Testbed 14 than I hoped for.
… This is the state of MapML at the moment.
… Why should it be standard? The barrier to use spatial information on the Web is currently to high. Requires programming, which is out of reach for many.
… Interest by developers including from JavaScript Map API developers (e.g. Leaflet)

ChrisLittle: You mentioned missing bits...

PeterRushforth: Two examples: A capability to modify/add the content -> form-processing. Feature content is unpredictable (e.g. elements within <properties> are unknown to a client) -> capability to add semantics using microdata.

jtandy: Why microdata?

PeterRushforth: Not so much microdata than HTML. Similar to itemscope/schema.org, leverage mainstream.

MichaelGordon: How extensible is MapML? JavaScript Map APIs allow extensions, can MapML support this and has hooks for such extensions?

PeterRushforth: MapML intends to be HTML for maps, so not extensible per se and somewhat rigid. Markup is content oriented, styling uses CSS.
… HTML is extensible using JavaSript, there yould be something similar in MapML where scripts could be used to provide extensibility.
… (shows MapML document)
… (Shows URL template mechanism - added in Testbed 13)

MichaelGordon: Is indexing by search engines an advantage? Probably more code for MapML than a Leaflet script required

PeterRushforth: Yes, but this is declarative and has a clearer semantics.
… Semantics of the map is obscured by the JavaScript, clearer in a media type

tidoust: Still unclear about the definition of "map". At one point a sequence of layers, in the MapML example layer does not exist.
… Related to this: Did you have contact with browser vendors?

PeterRushforth: Abstract concept of a map is 2d and supports mesh-up. The MapML example essentially is a map layer.
… The browser developer community is tough to interact. This proposal is a bit coming out of nowhere for them.
… The only way that maps become part of HTML is that the map community comes together.
… Browser developers look at adoption by web developers, but this is not about replacing the existing JavaScript Map APIs.

<MichaelGordon> scribenick MichaelGordon

RobSmith: 1) Accessible not just for programmers, which is a good selling point,though might be a hard sell for a development community that has already done work to solve the issue 2) you mentioned modification and deletion was something that you already had to do - for geotagging video for electronic maps we don't need to but for deletion IDs might be applicable, ChrisLittle also pointed us to SMIL

<MichaelGordon> RobSmith and PeterRushforth to take this discussion offline in interests of time

brinkwoman: not clear what the added value of this is?

<tidoust> [FWIW, note SVG animations is not widely supported across browsers (and will be dropped from a future version of Chrome)]
...would you encode spatial data in mapml? would there be actual coords in there and would that not become verbose
...would you also be able to do 3d maps with this having just seen CityJSON?

<jtandy> (MapBox vector tiles is one response to overly verbose vector content)

PeterRushforth: web maps are often raster based and we are sponsoring some work towards vector tiles
...they don't completely solve the issue of users putting maps on the web.
...vector data has some inherent links and that's why the vector model should be part of MapML

jtandy: flavour of where this is going and not right forum for technical discussion, in interests of time how does this progress?

<MichaelGordon> The team says goodbye and thank you to ClemensPortele

ScottSimmons: how do you see MapML fit in with OGC standards like WMTS, WFS etc?

PeterRushforth: already leveraging WMS under MapML (see demo)
...so won't replace OGC web services, and MapML won't be successful if it has to replace that information - its objective is to reuse where possible

Jtandy: definitely in the scope of spatial data on the web - focus on rendering it
...should be on the funnel

tidoust: already on the funnel (checking where it is)
...not in there

jtandy: should be in the funnel - in remit and could go to standards working group. tidoust: where does it fit?

tidoust: fits in incubation as community group but investigation because it needs a larger community.

PROPOSED: add MapML to the funnel at the incubation phase

<tidoust> +1

<jtandy> +1

<MichaelGordon> +1

<ChrisLittle> +1

<ScottSimmons> +1

<brinkwoman> +1

<billroberts> +1

Resolved: add MapML to the funnel at the incubation phase

tidoust: could MapML exist as custom extension not natively supported by browsers
...? we have other things like this musicml etc

jtandy: is there any point in developing it if there's no pathway to native browser support?

tidoust: goes back to brinkwoman's question on what the semantics are that are trying to be put on the web
...musicml looks to make music indexable etc as an extension

RobSmith: question of who is going to use it along the way as extension

tidoust: browser vendors are reluctant to add more media types

jtandy: seperate native browser implementation of map type, we're left with development with mapml encoding, Peter do you know how many other organisations that are supporting this and willing to create implementations? Seems like the biggest challenge

PeterRushforth: yes it does, could be used by various technologies - gis, leaflet etc without native browser support. In and of itself, it embeds map semantics into a hypertext

jtandy: do you know specifically who would support it?

robsmith: or are there specific use cases for it?

jtandy: do we have an understand of value add and who will say they will implement that?

PeterRushforth: could be adapted to simple leaflet as layer type and same could be said for openlayers
...it is open source so people need to be paid whilst implementing it

jtandy: if we want to progress to working group in w3c - we need 25 organisations to vote for this. Do you know of 25 orgs in w3c that would support this?

PeterRushforth: Not yet

jtandy: this group can help socialisaiton of the idea

robsmith: approach for doing that could be identifying the use cases and then approaching those users in that use case to show the potential benefits

jtandy: Peter do you have a similar set of use cases or user communities?

PeterRushforth: one of the objectives to lower the requirements on the HTML authors
...we want everyone to get value from our opendata. We want to lower the barrier

jtandy: we are trying to find a way the incremental value of this approach over existing ways, and then use that as message to persuade people
...that would be a useful discussion - how it is we can socialise the incremental value of mapml

PeterRushforth: Mapml community group is public so no barrier to joining
...both maps and HTML are both seen as having value, objective is bringing those together

jtandy: time to cut off due to time restrictions
...30 mins past finish time

tidoust: lets postpone that discussion

jtandy: is there a holding place for funnel ideas?

tidoust: Exploration space - anyone can put stuff there

jtandy: if we know of things that are pending such as map accessibility then lets create issues there so when we do plenary we can look at candidates
...tidoust are you able to create issues for those?

Action: tidoust to create candidate issues for funnel

jtandy: with that part of discussion postponed it would be useful to do an AOB. Ways forward with how we're organised, stats BP, spatial BP, SNN/SOSA errata and funnel. Anyone?

ScottSimmons: Critical at this time work queued up for work in either organisations, summary for ogc would be highly useful. Think we're making progress, would like to see more things come through and be assigned to OGC working groups to allow for scaling of work

jtandy: we've already said that we want to do a 5 min slot at Orlean TC - would that material be good start to discussion

ScottSimmons: yes and I'd allocate more to planning committee

jtandy: would you invite myself or brinkwoman to present at PC to attend for that discussion?
...the funnel will exist plus the presentation given to the TC

ScottSimmons: that would be good

jtandy: anything from w3c?

tidoust: happy with the progress and renewed enthusiasm in this group, funnels and roadmaps plus dinner on geonovum

Action: jtandy to present at Orleans PC

<trackbot> Created ACTION-384 - Present at orleans pc [on Jeremy Tandy - due 2018-02-27].

ChrisLittle: now we have a funnel - I could use that for a temporal ontology to look to progress this

Action: ChrisLittle to write a paragraph on this and put an issue in for temporal ontology

<trackbot> Created ACTION-385 - Write a paragraph on this and put an issue in for temporal ontology [on Chris Little - due 2018-02-27].

RobSmith: during presentation I showed the explanatory spec - whilst I would like to share it I'm not sure this is the right time or forum to share it. What are peoples thoughts?

jtandy: do RobSmith have a website for the spec itself?

RobSmith: only company homepage

jtandy: reticent to do this because it's hard to comment on. More than welcome to get a sub folder in interest group as a placeholder there

tidoust: or could create own github repo

jtandy: have a think about best place to do this not on company homepage

RobSmith: putting it under the IG does gives a focus on it

jtandy: RobSmith please could you work with tidoust to create the right folders

brinkwoman: just wanted to thank for group coming here

jtandy: thanks Geonovum and brinkwoman for wonderful hosting

<MichaelGordon> Jtandy closes the meeting

<ChrisLittle> bye

Summary of Action Items

  1. François to migrate the errata document to the GitHub repository
  2. tidoust to create candidate issues for funnel
  3. jtandy to present at Orleans PC
  4. ChrisLittle to write a paragraph on this and put an issue in for temporal ontology

Summary of Resolutions

  1. Re. SSN issues, idea is to: continue disussion, find resolutions on GitHub, collect solutions in a separate Note. Solutions may be implemented in a separate ontology files. Eventually this may lead to the chartering of a new SSN WG if there's enough material.
  2. Add Video Geotagging Protocol For Electronic Maps to the funnel
  3. Add CityJSON to the funnel in the incubation phase
  4. add MapML to the funnel at the incubation phase
Minutes formatted by Bert Bos's scribe.perl version 2.37 (2017/11/06 19:13:35), a reimplementation of David Booth's scribe.perl. See CVS log.