POIWG 2011 F2F #1, Day 1

29 Mar 2011


See also: IRC log


Matt, Christine, Fons, Alex, Jaques, Carsten, Jonathan, Dan, Ronald, Dirk, Luca
Andy, Gary, Karl, Raj
Alex, Matt
Luca, matt, Ronald


<trackbot> Date: 29 March 2011


<matt> Scribe: Luca

Round table

Fons kuijk new in Poi and he's interested in AR

Cristine Parey Invited experted

<matt> [[New participants: Fons from W3C Office and CWI, Carsten from Ordnance Survey and OGC]]

Jonatan from ETRI in Korea

Dirk is CTO of Layar hosting the meeting

Dirk: thank you all to coming

<cperey> Luca Lamorte, Telecom Italia, AR interest

<danbri> i'm Dan Brickley <http://danbri.org/>, occasional geo-geek, semi-lapsed WG member.


Matt: first talk about the agenda
... talk about talk about metadata stuff
... topic q

<matt> Agenda

sorry topic 1

please look at the wiki page for the agenda

<JonathanJ> Agenda: http://www.w3.org/2010/POI/wiki/Face_to_Face_Meetings/March_2011

matt: How we going to plan the draft
... Talk about next F2F
... then we have the OMA liaison request

<danbri> dan: I'm in the new RDF Working Group ... we have a deliverable of a JSON format for RDF. Group is currently trying to decide requirements for that, ... so if this POI WG has JSON needs, that could be good input.

<danbri> ... happy to be a liason

matt: thanks Dan

Dan: Use to JSON and care of RDF

<danbri> ACTION: danbri circulate pointer to RDF WG JSON discussion [recorded in http://www.w3.org/2011/03/29-poiwg-minutes.html#action01]

<trackbot> Created ACTION-36 - Circulate pointer to RDF WG JSON discussion [on Dan Brickley - due 2011-04-05].

Matt: AR vocabulary
... Any changes of agenda?

<matt> Results

Planning First Public Working Draft

<matt> Charter

Matt: first draft called CORE it should be ready in April
... late april, early may it's not so bad

<JonathanJ> W3C Process document : http://www.w3.org/2005/10/Process-20051014/tr.html#first-wd

Matt: how we want to go on the document?
... we need so cohesion
... a role of editor
... the author is the working group
... Editor should put on the proper format
... anyone want to step forward?
... I can put a strong effort help on this issue
... the editor is also the author
... and we can work as we want

Alex: how we can expect people to work collaboratively without clashes, etc?

Matt: the social part
... we can have hierarchy
... so far people is scared on put things on the wiki
... they prefer to send stuff by email, then someone adds it to the wiki


<JonathanJ> we can also use to tracker system - http://www.w3.org/2010/POI/track/

Carsten: it's good put stuff on the wiki and then stop a certain point
... to make an agreement and edit them into the document

Fons: it should be a native speaker of doing it

cpery: there is not so match traffic on the mailing list

matt: we need to put more action points on the call, so it would be more effective

<matt> POI Tracker

<matt> RESOLUTION: Matt to start acting as informal editor. WG will bring together issues weekly for entering into wiki.

<danbri> draft: http://www.w3.org/2010/POI/wiki/Core/Draft

<matt> Web App Tracker

<JonathanJ> one of example from MWBP WG (closed WG) - http://www.w3.org/2005/MWI/BPWG/Group/track/

matt: They have raised 46 issues

and only 6 are closed



<matt> MWBP issue

<danbri> http://www.w3.org/2005/MWI/BPWG/Group/track/issues/

matt: look at one the issue
... issue can also have actions associated on that
... back to irc
... Move on the the next topic?

Requirements Generation

<danbri> what do we have re requirements?

<danbri> http://www.w3.org/2010/POI/wiki/index.php?title=Special%3ASearch&search=requirements&go=Go

Dan: can we look at the requirements?

<matt> Use Cases

matt: i start put some requirements


matt: I collect tech requirements here http://www.w3.org/2010/POI/wiki/Requirements
... maybe we should go back into primitive

Alex: would say personally yes because I think we don't have finished the discission
... we need a review

<danbri> I'm looking for stuff like: "I maintain a Web page listing points of interest in my city; what could/should I do to make this W3C POI compatible?"; "I'm building an AR app for iphone, what data structures should I create to hold W3C POI data?", "I'm building a database of POIs to back a social Web site, ... what minimum structure should it have to fit the W3C spec?", "I have a javascript library for plotting POIs on a 3D globe - https://mozilla

<danbri> demos.org/demos/globetweeter/demo.html eg - ) ....etc"

Alex: My concerns about relevance, we need a short review on what we have said

<matt> Current Use Cases

Alex: these use cases sounds to me like are almost the same

<JonathanJ> https://mozillademos.org/demos/globetweeter/demo.html

<matt> Christine's use cases

alex: I hope there are use case that covers all and not only one aspect

cperey: some agend are creating new content

alex: people have practical thing to do in their head and these has effect on the UC
... we need to spend more time on UC, because we got away from the story

Carlsten: we need to define what is out of scope

cperey: maybe some of the issues is that
... OGC are more familiar on people who are

<danbri> no

<danbri> (matt, sure. notes here:) ... 2 proposals to make (test driving Issue-based discussion and decision making), (i) POI spec should make explicit ability to distinguish an organization (eg. a school) from the building/geo it inhabits (ii) POI spec should make explicit ability to distinguish organizational brands (Starbucks, CostaCoffee). Suggest we decide this in terms of personas for relevant use cases, eg. webdev working at starbucks; 3rd party

<danbri> layar developer making a 'coffee hangouts' layer or iphone app....

<danbri> (if we can spend a little time to walk thru discussing those and making or not making issues, that'd be lovely)

Alex: we need more people experts on different aspect we want to cover on POI

danbri: 2 expample on modelling issues

sorry too fast to me :(

<matt> scribe: matt

danbri: 1. Are we going to have explicit ability to distinguish between say an organization and it's buildings?
... 2. Distinctions between organizational brands, e.g. all of the Starbucks in Boston.
... I don't think we have enough use cases to cover these.
... I think we should have more personas, say GIS professional or AR developer
... Say, I've spent time building a layer of Amsterdam restaurants, how does that person relate to the spec?

-> http://www.w3.org/2010/POI/track/issues/new Create a new issue

<danbri> http://www.w3.org/2010/POI/track/issues/2

-> http://www.w3.org/2010/POI/track/issues/2 organizations vs places

[[WG works through capturing issues in tracker]]

ahill: Let's not necessarily flesh it out now, but log it and get us into the habit of raising issues.

-> http://colloquy.info/ Colloquy Mac client

<ahill> rrs agent draft minutes

<JonathanJ> help about rrsagent - http://www.w3.org/2002/03/RRSAgent

<Luca> next f2f

Next F2F

<scribe> scribe: Luca

<JonathanJ> manual for zakim - http://www.w3.org/2001/12/zakim-irc-bot.html

matt: hosting the meeting ...

<danbri> ( re above discussion see http://www.w3.org/2010/POI/track/issues/3 poi-brands --- and a bit of discussion re process )

matt: 2/3 per year
... big meeting in novemeber
... in Santa Clara

<matt> TPAC 2011

matt: So from October 31st
... there is a plenary where all WGs are in the same room
... and each groups talk about what's going on

<matt> ->http ://www.ismar11.org/ ISMAR 2011

cperey: the OGC has a meeting in september

<matt> OGC Technical Plenary

cperey: it's better to avoid meeting while there are other at the same time in other place
... that was one the problem during Atlanta's one
... The next one should be in ASIA

alex: what does meaning hosting rotating?

matt: To be fair to all participants we should try to rotate through the continents they are located in.

cperey: we don't have ideal time for meeting on each continent

<JonathanJ> we can host a meeting in seoul

<JonathanJ> a few weeks ago, we hosted DAP WG meeting in Seoul - http://www.w3.org/2009/dap/wiki/SeoulF2F2011

matt: we should have one meeting in summer

cperey: also OGC has volunteered for hosting
... would like to OGC members to participate
... OGC will pay the place/food/registrstion

<danbri> re OpenStreetMap I find http://wiki.openstreetmap.org/wiki/State_Of_The_Map_2011/Bid/Denver but not the final url

Alex: how we connect organization like W3C with OpenStreepMap
... Just a practical question

<JonathanJ> If we can host a meeting in seoul, potential Korean company(LG and Samsung) and Korean AR guys could be join us.

Resume proposal on WIKI


<danbri> proposal maybe? "POI Core format should be capable of round-tripping Open StreetMap data for points, potentially using extension mechanisms to capture any extra info needed. A draft CR exit criteria?"

<danbri> OSM data: http://wiki.openstreetmap.org/wiki/Planet.osm

alex: how POI describe OpenStreetMap

<matt> PROPOSED RESOLUTION: Meet face to face 3 times this year

<matt> http://www.w3.org/2010/POI/track/issues/4

<matt> Scribe: matt

cperey: There was a two day meeting after Barcelona. This group wasn't represented.
... Alex gave a presentation, Andy was planning to come but couldn't make it.
... OGC proposed to make itself the administrative platform for a loose community of those who care about standards.
... George Percival made a presentation and a pitch about it.
... I've had a few meetings of international standards community that have brought together 50-60 people. Half presentation half discussion.
... One proposal is that the OGC take the lead.
... There has been some pushback. No decision has been made.
... We didn't have visibility into other standards organizations.

<danbri> (this all about AR...)

cperey: It's informal, grassroots, people can contribute, but since it's informal, it doesn't get official recognition.

danbri: Similarly WHATWG made things and made people listen.

cperey: There's a very high quality discussion on the list. It's very AR focused, not focused on one specific domain.
... So proposal for f2f to be at the next OGC meeting, I went to George and said "we can't just continue to do the same formula", need to have concrete objectives. Can't just say we'll chat about position papers.

<danbri> see also http://lists.w3.org/Archives/Public/public-poiwg/2011Jan/0021.html http://arstandards.org/mailman/listinfo/discussion http://arstandards.org/pipermail/discussion/

cperey: So the accepted proposal was that OGC set up liaison between OGC and other organizations and then a meeting outside the OGC, where others can attend who weren't members.

<danbri> lots of discussions here http://arstandards.org/pipermail/discussion/2011-March/date.html

cperey: It would let multiple organizations get together.
... The other option is to have liaisions.
... To short-circuit things, we would be to create liaisons with the other groups that skirt at the ages of where we are.
... The proposal is to have a multi-SDO meeting. Maybe people are not part of each organization.
... The purpose would be to create a figure that defines what different organizations, be they grass roots, or more established organizations, that describes who is going to take care of what.
... So the scope is understood between one another.
... Figure out our constituents, and then let the other parts that have expertise be covered by other organizations.
... So if this working group could attend one day, or half a day, and have it's WG meeting around it, that would be one proposal.
... Each SDO present what they are working on.
... It would really accelerate the work on the AR landscape.
... It would in one place bring in where all these organizations are.
... I've written a chapter for a book on the subject of where we stand now as far as the standards organizations are concerned. Co-authored with Carl Reed (OGC) and Timo ?? (Fraunhoffer).
... and a contribution from the Khronos group.
... We created a diagram of which groups have which expertise. I'm happy to share it.

alex: So there's a proposal on the table related to what you first mentioned. This invitation by OGC to go to Taiwan in June. Does this group want to have it's F2F there and participate in a multi-SDO meeting.

cperey: Those that join OGC covers the participation in the TP, etc.

carsten: From OGC there are two types of WGs, there's WGs which are difficult to liaise with, and then there are more general domain WGs.

cperey: Examples? Mass Market?

cartsten: And 3D Information Management

danbri: And GeoSPARQL?

carsten: That one is a standards WG, but there's also geo semantics, which is a domain WG.
... 3DIM covers an awful lot of stuff that may be relevant in an AR world ranging from geology to navigation of robots within buildings.

<JonathanJ> one of OGC's activity in W3C was SSN IG - http://www.w3.org/2005/Incubator/ssn/

carsten: I'm not sure there will be much 3D stuff in Taiwan, so I think the Boulder meeting might be more relevant.

matt: The Boulder one also has the OSM meeting around it.

carsten: There may also be a 3D summit in Boulder.

-> http://www.w3.org/2010/POI/wiki/Face_to_Face_Meetings/2011_Future_Face_to_Face_Meetings#2011_F2F_Options

<danbri> looking for public OGC discussions -> http://www.ogcnetwork.net/forum/ ... is that a useful place to hang around?

<danbri> searching 'geosparql' i find 3 links: http://www.ogcnetwork.net/search/node/geosparql

<danbri> googled -> http://www.opengeospatial.org/projects/groups/geosparqlswg/

[[insert matt's comments here]]

alex: I think you're saying the liaisons are sufficent, but getting everyone meeting is not necessarily the solution to the problems.

matt: I think we can get the actual work done best through liaisons. We can use the official W3C workshop mechanism to do the rest, or we can have another group do it, or a new third party.

cperey: I think the struggle we're having on the list is a reflection of a larger question.
... I don't believe it would be healthy to have a multi-SDO meeting that is a mock-up of having five standards development organizations there, and only have 2 show up.
... If the WGs within the SDOs, the work item groups, the OMA effort, etc, decide to come and OGC is there and W3C and Khronos is there, if you can get 4 of 5 or 5 of 6, that tips the scales to looks promising.
... Then if those organizations have their liaisons in place and can make their activities more known to one another, then that would be beneficial to all of the work.
... There's a lot to be hashed out, it's not a done deal that this meeting will happen. It's voluntary.

<danbri> no sweat

alex: I hear that if we don't liaise with these other groups, then we fail in a sense. I think that's a real problem, I don't know if what you're proposing is the solution, or if there are other ways we can liaise and succeed.

danbri: One of the biggest obstacles is that some of these organizations work in private. I think the tendency is to work fully in public at w3c now. It was hard. Every time you move stuff public, you remove the incentive to join. OGC, Khronos, ISO all have these problems.

<JonathanJ> +1

danbri: Someone has to pay for making standards. What we have with AR and POI is a very broad, loose, hard to get your head around topics.
... 3d modeling, QR codes, etc everything is in there. Because it's so broad, I think having a single meeting with all of the stake holders can be useful.
... We need to have as broad a group as possible to send links and share.
... Public discussions, things that have links... if we don't have that we have press releases from CEOs talking about working with other groups and nothing inside.
... I think we should be moving more and more public in this WG.

alex: Do you think much happens in AR, e.g. the Layars of the world, are talking in public or in private about standards.

danbri: They're busy getting started. As a market leader, it's difficult to rationalize standards.
... Some standards yes, some no.

<JonathanJ> Another biggest obstacle is IPR problem.

Ronald: I think the AR browsers have realized there will be a standard, and we can't wait for it, so we're doing our own stuff, so we want to be the first to make a browser for that standard too!

cperey: Layar and Mobilizy contribute an enormous amount to this subject.

alex: I think I narrowed it down too much, but there's also the Nokia/NAVTEQ's that have interest in seeing these standards go forward, but I doubt they are talking to each other a lot about those standards.

cperey: MetaIO came to the meeting in Barcelona and was an eye opener for them.

<danbri> (absolutely no disrespect to layar intended; i just mean that it's expensive to put time into standards, but for market leader the incentives are always ambiguous for std'izing a rapidly evolving platform)

alex: Yes, it's obvious that the Wikitudes and the Layars of the world see this as a good place to touch. The question is what organization do we create that is appealing to them.

<danbri> much for nametags :)

alex: There will be some organizations who don't see it as their full time job to be in standards organizations.
... It sounds like it's similar to OGC and W3C as it has membership, IP, companies have things to say in it.

cperey: And OMA.
... They don't want to have membership in five organizations. They want to have confidence that the organizations will define the interestions.

alex: Do we want this to be a permanent organization?

<JonathanJ> I worry about that Giant company(Google, Apple, MS) can be a killer of AR standardization in future.

cperey: Which we?

alex: All of us at the same time.
... If they don't want to be in each organization than they can have their own group, or they can join OMA, or...

<danbri> I can't get excited about another group to liaise thru/with/via, as a solution to the 'there are too many groups to liaise with' problem :|

cperey: It can be counter productive to be too formal.

alex: On one hand, we're not a bunch of hackers throwing together all of our AR fantasies. We're trying to be a little bit more "pointed". We can't have both at the same time.

<scribe> Scribe: Ronald

<matt> RESOLUTION: WG will meet f2f three times per year

<matt> Future F2F meetings

Matt: recap meeting options
... as we want to meet 3 times this year, we have a good starting point for selecting meeting locations
... can be decided later on during our calls
... do we want to continue with the liason discussions, or move on to another agenda item?


<matt> OMA Liaison statement

Matt: lets start with the OMA Liaison for discussion
... OMA has 3 specified deliverables, data format, transport client features and security features
... Clear overlap with data format

<JonathanJ> Here is initial draft of OMA MobileAR Enabler specification - http://db.tt/JmtFaNO

Luca: not personally involved in OMA, but Carmen, which is in the same group at Telecom Italia is in the OMA group

Jonathan: I am in the OMA MobileAR meeting in two weeks
... not involved in the previous group calls
... Samsung joined OMA, but not in the MobileAR Enabler

<matt> Luca: Supporting companies: EnSoft, LG Electronics, AT&T, Danal Entertainment, Telecom Italia, Samsung

<matt> Luca: they have different membership types.

cperey: qualcomm is active as well

matt: we have people who are members in both organization, we have an agreement with OMA and OGC for data sharing

<JonathanJ> I have asked Criminisi (Current Champion of MobileAR), Telecom Italia, but she said not yet decide for MobileAR meeting in Sorrent.

ahill: is there any reason against it?

matt: as our work is public, there is no real need for it

cperey: seems like a no brainer to do a liaison
... having people in both groups is not a real way of knowing what is going on, but having an official liaison is better

matt: we should talk and be in cooperation and point out where the overlap is with us and other

cperey: do we have POI covered? We are having problems getting involvement

ahill: it might make other members from OMA to participate in our group
... agreeing to liase with OMA does not influence our charter

matt: OMA wants to write a standard for POI. we should talk to them and work together instead of duplicating efforts

cperey: creating a contact surface of POI for immediate discussion with them. Ask them how they want to define POI

matt: having them at the table is ok.. drafting the response has to be looked at by a lawyer from the W3C perspective

luca: POI is more general than AR
... we have to look at how it will affect our groups work

<JonathanJ> As you can see in Mobile AR spec - http://db.tt/JmtFaNO , I think they are just focused on Mobile AR architecture.

luca: the OMA work package seems to be describing an application and not really a specification
... we don't have a special interest in AR, but have it as a use case
... they will need us as a common way of information and we can take their specific use case as input for our POI definition (e.g. special data field requirements)

matt: our generic POI is supposed to be extensible, so if OMA want a standardized client for AR they might want to add specifics to the data model

cperey: the scope of OMA seems quite large, it covers the entire stack
... we have taken only a small portion of the problem and it is already messy.
... the companies involved are not the ones who are currently working on it in the real world

ahill: maybe its the nature of OMA that they want to standardize bigger things and this is just a blanket

matt: the m stands for mobile, so they limit their full stack to just mobile ar
... they are in a strange place compared to us

ahill: they are not just doing it in private, but they are seeking liaisons. so maybe they realize they do not have all the answers

<matt> ACTION: andy to start drafting a response to OMA liaison statement to agree with the idea of the liaison and to suggest that they participate in our POI and AR discussion and try to converge their technical specifications with ours [recorded in http://www.w3.org/2011/03/29-poiwg-minutes.html#action02]

<trackbot> Created ACTION-37 - Start drafting a response to OMA liaison statement to agree with the idea of the liaison and to suggest that they participate in our POI and AR discussion and try to converge their technical specifications with ours [on Andrew Braun - due 2011-04-05].

<danbri> JonathanJ, thanks for that link, I hadn't seen it. There's a basic taxonomy of actors in OMA-ER-MobAR-V1_0-20110311-D_20110323 but it seems very simple (user, content provider, service provider)...

matt: when looking at Khronos, we don't want to specify 3D specs, and we don't want them to deal with POIs

ahill: not sure what Khronos would contribute to POIs, but it would be nice to know they are involved
... it will help our legitimacy
... maybe we should shift to a more tangible look at what the core draft is to summarize

Reviewing the Core draft

<danbri> http://www.w3.org/2010/POI/wiki/Core/Draft

ahill: the core draft we have is the result of our F2F in december
... discussions were about whether POIs were purely geo based or describing things
... a core POI is described by a location, name, some relationships, identifier, category, some temporal information, and any metadata, any extensibility
... the first one is location
... we wanted to have something flexible, not just individual points
... it can have one or more location primitives, for example 100ft from the corner of another known place
... or some strict WGS84 location
... or a description for an image capturing device, so you don't know the real location, but do know it when you see it
... if some service exists that can resolve the exact location that is helpful
... we do want to look at 3D objects for extends
... we did not go into detail
... the relationship primitive is for defining relations for example adjacency, franchise of a chain,

<matt> Carsten: Silicon Valley is a "vernacular name"

ahill: for navigation adjacency is important to have

cperey: if you have 5 stores in a mall, how do you specify that the specific POI is in the mall

ahill: here we got a bit confused. we have extensibility, relationships, etc... and we did not know what should fit in which primitive
... the category has to be added to POI from the big POI databases perspective

cperey: the categorization pushes us into the direction of linked data

carsten: it is an area where the data vendor pushes their picture of the world

danbri: we know we as a group cannot specify all categories, but leaving a giant hole in the spec is not good
... so we could specify our standard to let others insert categories

<danbri> (there might also be different kinds of categories; eg. cultural heritage categories for historical landmarks, versus types of company, ...)

carsten: in citygml we try to model the real world, so we have materials for building, roofs, which can be country specific.
... initially, we wanted to give an example in the specification and let people use their own

<danbri> (another example, friend here in AMS collecting info about road blockages for fire dept usage ... categories range from 'big hole in street' to 'lighbulb needs changing')

carsten: but if you publish the data, it is not that useful. so we looked into having a registry for "approved" lists
... trying to convince OGC staff to look at a resource to utilize, but still at the beginning

<danbri> (q for Carsten, I heard a rumour OGC was moving towards using SKOS for 'code list' categories?)

ahill: I am still new to the RDF world, but there are some standards. You would not create the registry ad hoc

carsten: there is a bit of structure. we are currently in the process to put an URI on top of it to find it

<danbri> http://www.google.com/search?sourceid=chrome&ie=UTF-8&q=skos+ogc

ahill: isn't there something in the RDF world, like a schema?

danbri: skos is more of less a thesaurus with different levels of categories

ahill: my point was that when people go to registries, it should not be ad hoc. we have some guidance

matt: with RDF you can look at the namespaces and dereference

<JonathanJ> SKOS - http://www.w3.org/TR/#tr_SKOS

jacques: perhaps we should focus on XML first as it is simple and look at RDF later
... RDF can be powerful, but we can stay with something simple and extensible

ahill: if I see something from "good relations" but how do I know what the document contains

danbri: you can look at the spec. It is not always writing the code for you
... it is semi-structured chaos\

ahill: it is up to me if I see foaf to know what to do with it

<danbri> oh this seems useful: https://portal.opengeospatial.org/files/?artifact_id=29334

<matt> Jacques' message

matt: in januari, jaques sent a sample of RDF that uses foaf an poap

ahill: there will be a lot of data in a POI, but how do we define what data we want to get
... I am assuming linked data has a solution for this

danbri: there is a linked data query language SparQL

ahill: I was expecting something easier

<matt> GeoSPARQL editor's draft from OGC

ahill: I don't necessarily need all the information about ticket prices if I walk by a movie theater
... is there a way to hit a URL to get specific data

<matt> Paper from Geodata

matt: jacques provided a link today

<danbri> i'll try converting that POAP example to fit into Facebook OG format, as a quick demo

<matt> Ronald: The POI information is at first limited, but you can view in a web view more information

<matt> Ronald: Kept the APIs simple and deliberately chose to have a single API call.

ahill: it is a big problem for anyone building mobile AR
... might be our of scope for this group

matt: It is in scope to make something useable

jacques: we have an app build on RDF, which is powerful, but we don't use a lot of it
... not sure if companies like Layar, Wikitude, Metaio would leverage the power of RDF

matt: we decided last F2F to not look at the technical format, but primarily at the data model
... and have multiple options for data representations

danbri: if we have an entity-relationship model we are pretty close to RDF anyway\

jacques: when looking at openstreetmap, it is very popular because the XML format is very simple
... we now see many people converting the data to RDF format
... in this group it might be bettor to have a simple XML format. Others can map it to RDF

<matt> OSM format?

danbri: the data will appear in multiple places JSON for web, XML, etc
... whatever we do we will upset people\

jacques: authoring json is more difficult than xml
... openstreetmap has 2 billion POIs in XML, with semantic data in each POI
... which is why I prefer XML, it is like HTML

<JonathanJ> Comparison of JSON and XML Data Interchange Formats: A Case Study - http://www.cs.montana.edu/izurieta/pubs/caine2009.pdf

<JonathanJ> another comparison - http://www.scriptol.com/ajax/json-xml.php

<matt> scribe: Matt

Ronald: I'm not sure we should focus our efforts on the mapping, but rather on the datamodel.
... We should choose one format for our internal discussion and focus on that, rather than focusing on all of the mappings at once. Let's look at the data model and pick one format, say, XML for doing examples, etc. We can keep an eye on the mapping, but not try to do the exercise of mapping it as we go.
... We do everything in JSON at Layar, but I can understand why you wouldn't want to, for the fact that you'd invent namespaces on top of JSON, etc.

Alex: I have this sneaky feeling that we're going to end up with linkages between location entity, or any other sort of data, be it a sound or a description or a 3d model. And that sounds a lot like Linked Data and RDF.
... As long as the XML that we are writing can go to RDF, that seems right.

Ronald: I agree. I see a real future in this linked data stuff.

Alex: I like the idea of describing it in XML to not overwhelm people.

PROPOSED RESOLUTION: Group will work in XML, and not do the mappings to other formats simultaneously, but keep mappings in mind as we go.

Alex: I'd like to talk to more people in Linked Data, and make sure we're not missing something.

<danbri> proposal: The WG expect that broadly equivalent POI descriptions will be exchanged in the Web using at least XML, HTML5, JSON formats,

<danbri> often embedded within some surrounding format or spec. For example, POI details might be described within HTML markup inside KML XML, inside

<danbri> Atom, as part of OpenGraphProtocol RDFa in HTML, or in JSON and JSONP Web APIs. The value of the WG's spec's will in large part be related to

<danbri> their ability to communicate POI data across such application boundaries, especially where extension information (eg. categories, relationships)

<danbri> needs to be preserved.

<JonathanJ> +1

matt: +1

<Ronald> +1

<danbri> (nobody objected :)

PROPOSED RESOLUTION: The WG expect that broadly equivalent POI descriptions will be exchanged in the Web using at least XML, HTML5, JSON formats, often embedded within some surrounding format or spec.

<danbri> .... in fact http://www.w3.org/2010/POI/wiki/Core/Draft already goes a nice long way towards identifying those commonalities

<scribe> scribe: Ronald

RESOLUTION: We should have just one normative mapping to use in the document

<danbri> matt wants at least one concrete doc format that can be canonical (at least within group's own discussions)

<matt> RESOLUTION: The WG expect that broadly equivalent POI descriptions will be exchanged in the Web using at least XML, HTML5, JSON formats, often embedded within some surrounding format or spec.

<danbri> (I don't mind which format so long as we've got a pluralist deployment story)

<matt> RESOLUTION: The normative mapping included will be XML.

danbri: in the end back end developers will use their own representation based on the abstract model
... they need to be able to read the spec
... don't assume people will have xml databases

ahill: let's continue with the summary of the core draft
... the time primitive, we still did not define fully
... question on whether it reflects the time when the washington monument was created or the time when the POI was created in the system
... a POI must have a location primitive. From geo, a POI only exists when it has a location. From AR, we might not know where the POI really is

carsten: a location isn't necessarily a coordinate
... we deal a lot with postalcode information, point information for delivery
... when a new postalcode is created, we don't know the exact point yet
... attribute indicating the POI is unmatched
... it is in this mall is a valid location

matt: we talked about having an "unknown" location, but we can probably have an incredibly inaccurate location

ahill: there is a segway to relationships
... we have in the spec "undetermined" for location
... but there are relationships that can then determine the location, e.g. a relation to the mall POI
... there is always a location primitive
... still need to clarify things a bit more

<danbri> (where's zakim gone?)

ahill: a lot of things are still to be decided

<matt> Relationship Primitive

danbri: last year I defined a list of possible POIs. A lot of them were event based. Are there relationships between events and places

<matt> Danbri's list of potential POIs

Ronald: is an event a POI or is it temporal data for the venue POI

ahill: maybe this is time for an issue

<danbri> Q: Do we deal with event descriptions (eg. movie screenings) by making them POIs, or by describing their relationship to POIs?

<danbri> 16:10 danbri: Q: Do we deal with event descriptions (eg. movie screenings) by making them POIs, or by describing their relationship to POIs?

<matt> ISSUE-5?

<trackbot> ISSUE-5 -- Do we deal with event descriptions (eg. movie screenings) by making them POIs, or by describing their relationship to POIs? -- raised

<trackbot> http://www.w3.org/2010/POI/track/issues/5

<danbri> (my dayjob is TV recommendations, so I have some interest in this use case :)

<ahill3> aSD

<ahill3> ASDF

carsten: do you guys think a POI is an address?
... are we thinking in 3D or 4D?
... an address can be streetname and house number, or lower level numbers specifying a location

<ahill3> what IRC server address in limechat?

<danbri> so I copied jacques RDF example into github repo -> https://github.com/danbri/Pogo/tree/master/checker/testcases/poi .... made versions in 3 different RDF syntaxes: Turtle, N-Triples and RDF/XML

carsten: if we are thinking 3D, an event is probably not a POI, if we are thinking 4D it might be

<JonathanJ> In Korea, we are developing the ARI(Augmented Reality Identifier) spec that it extend to IETF's URN specification.

carsten: people tend to not look at the temporal aspect and just create "static" POIs
... we have the chance to boost the time aspect

ahill: if you don't add a time element to a POI it is seen as indefinite/static

matt: the temporal stuff we definitely have to address while crafting the core draft

<danbri> proposal re temporal aspects: POI spec should allow for temporal characteristics of the description/record (publication date, 'coverage' / when it's true dates); and also of the payload, including domain specific extensions such as shop opening hours, movie show times, or historical periods for sites of touristic interest.

ahill: a POI can have contained within or contains, adjecent to relations
... they are defined between POIs
... must have a name primitive
... must have an id primitive
... lots of debate. Does a location have an id and are we reusing locations when two POIs share a location?
... same for categories. Does a category need an ID
... does a POI have a single ID, or do all the element it contains have IDs that can be referenced

<danbri> proposal2: any entities of interest should be able to have IDs, and these should be allowed to be http:-based URIs

ahill: must have a categorization primitive. Everything has to be categorized otherwise it has no value (from the Navteq perspective)
... open to debate, but it is in the current draft to trigger more discussions and not forget about it

danbri: in RDF there is no must at all. It is a very passive environment and you have to deal with what you have
... we can have best practice notes, but I feel unconfortable having it in the core spec as musts
... The POI spec needs to have a temporal aspect on the POI data as well as on the payload (opening hours)

ahill: there are 3 levels: opening hours, when did the restaurant open, and when was the POI created

danbri: if we have ids on everything, we can offload the temporal data to others

<danbri> -coffee needed-

<matt> scratch pad for us to work on

<danbri> test case fodder: http://www.pathe.nl/bioscoop/tuschinski http://cinematreasures.org/theater/2314/ http://en.wikipedia.org/wiki/Tuschinski http://dbpedia.org/page/Tuschinski http://www.cinemacontext.nl/id/B000030

<danbri> and a movie showing http://www.pathe.nl/film/7717/the-kings-speech

<danbri> times: http://www.pathe.nl/film/7717/the-kings-speech/agenda

<danbri> ...

<matt> Scribe: Matt

IDs, Microformats, RDFa, Open Graph Protocol and Linked Data

Alex: Michael recommended URIs for us to use as IDs. How do we use those?

danbri: If you are pointing to things anywhere on the Web, the URI spec is the way. If you are looking at things within the document, you might just have a local ID, and then a convention for --

alex: I hit this URI and get back XML?

danbri: Well, broken HTML.

alex: I mean, how does this jive with XML?
... I can imagine that I click on this link and it returns XML for the King's Speech, but is that how an ID works?
... I'm thinking more like an XML ID attribute.

danbri: Within an XML document, depending on the format type, there are different conventions for how you find these references. XHTML, SVG, etc will all have their own.

alex: What is the role of IDs?
... A very simple standpoint, for example in KML. You can identify things within the scope of your document.
... We had a proposal that URIs are the IDs.
... Are they IDs that were created by the author for internal use, or URIs?

jacques: In OSM, it's unique for each node.

Carsten: We've had that for the last ten years, we issue our own numbers and everyone using our data knew about them and it was accepted. We're not going to change this particular product any time soon. Geo data has national bodies in the UK.
... You want to use the data and mash it up with those other sets out there, so URIs are obvious.

<JonathanJ> Can we maintain centralized and universal unique ID ?

danbri: There are two parts: within an XML document, how do you identify things within itself and how do you point out?
... I can write "id='foo'" and someone can, on the Web compose a URI from where the URI where the document lives, plus say a hash mark and the ID.
... There's been a tendency to do it different in each spec, but XML ID spec gives guidance.

<JonathanJ> XML ID spec http://www.w3.org/TR/xml-id/ , XML Base spec: http://www.w3.org/TR/xmlbase/

matt: We would take the URI used to fetch the XML doc itself and combine it with the ID specified to give a unique ID to part of a given POI.

alex: I'm a bit concerned that we think URIs are good, but the details we don't get.

danbri: It's a bit fiddly.
... The linked data people tend to get excited about using HTTP URIs to name things. Tim put a note out saying use http identifiers to identify real things. People are excited about it, but I think we can be agnostic about it.
... Tim said "we can use the hash to make it not be web pages", but in the linked data context they can be used to reference people, tables, etc.

<JonathanJ> I think we are also need URI like ID scheme

Alex: if I have a hash in a URI, who sees it?

matt: It depends on the client. A web browser won't send it.

danbri: Or you get back a bunch and then look through it to find it.

alex: But they could take the hash part out and search for the id itself.

danbri: People take care to make sure that before the hash is enough to be able to pick out the single entity.
... There was a ruling about this in the TAG. They said it's okay to name a real world thing in a URI, but as long as it sends 303 redirect.

<danbri> 1. http://en.wikipedia.org/wiki/Tuschinski

-> http://www.w3.org/2001/tag/group/track/issues/14 HTTP-Range 14 TAG issue

<danbri> 2. http://dbpedia.org/page/Tuschinski

<danbri> 3. http://dbpedia.org/resource/Tuschinski

danbri: So, there's an HTML page that describes Tuschinski, then there is a DBpedia page that describes the page, and then a third one /resource/ that should give a 303

[[$ curl -D- http://dbpedia.org/resource/Tuschinski

HTTP/1.1 303 See Other]]

<danbri> curl -I http://dbpedia.org/resource/Tuschinski

<danbri> HTTP/1.1 303 See Other

[[Location: http://dbpedia.org/page/Tuschinski]]

danbri: So, this does several things, it maintains a distinction between the thing you are talking about and the description of it.

<Carsten> documentation about distinction between resource and page at location.defra.gov.uk/wp.../Designing_URI_Sets_for_Location-Ver0.5.pdf

danbri: There's a tendency to try to use URNs for real world objects, but you can't drop those into a Web browser and get something back.
... isbn://12312344 is a URN for example

<Carsten> correction: http://location.defra.gov.uk/wp-content/uploads/2010/04/Designing_URI_Sets_for_Location-Ver0.5.pdf

<JonathanJ> URI scheme - http://en.wikipedia.org/wiki/URI_scheme

danbri: What we see on dbpedia is one of the two main paths you can take for identifying a real world object. Either "#blah" or have this HTTP redirect dance.
... In which case you have something that looks like a URI link but you only find out by fetching it.

[[<MYPOIFORMAT id="tushinski">

<name> bla blah</name>



matt: This would be served out of http://example.com/tushinski

<danbri> so the full URI could be http://example.com/tushinski#tushinski

danbri: There is a cinema in Amsterdam, it's name is Tushinski, built in 1921.

<danbri> http://en.wikipedia.org/wiki/Tuschinski

danbri: It is described by Wikipedia in a document whose URI is:

<danbri> http://www.pathe.nl/bioscoop/tuschinski

-> http://en.wikipedia.org/wiki/Tuschinski Tuschinski Wikipedia Document

danbri: There is also a web page published by the film corporation:

-> http://www.pathe.nl/bioscoop/tuschinski Tushinksi web page

danbri: You can watch movies there.

Carsten: Is the home page about the movie, or the building?

Ronald: The home page is about the company in the building.

danbri: Going to the homepage, I get an embedded map showing where the theater is and others from the same company.

Carsten: Do we see the POI?

danbri: This is a nice example of pointless information barriers. Getting this into a navigation system is... fiddly.

matt: And how would we make it not fiddly?

danbri: Markup in the page using the POI standard could make a machine readable description of where it is, etc.

ahill: I like that.

<danbri> <meta name="google-site-verification" content="ewxq0ABm0oKTxJXMgwFYF3zVYTAH7iFbJZeQvpqY-Qw" />

<danbri> <meta name="keywords" content="Pathe, bioscoop, film, amsterdam, rotterdam, den haag, eindhoven, helmond, groningen, utrecht, munt, arena, schouwburgplein, utrecht, rembrandt, metropole, filmagenda, bioscoop agenda, filmfocus, filmtheater, arthouse, uitgaan, belbios, nederland bioscoop agenda, nederland bioscoop, recensies, films, prijsvraag" />

matt: RDF also has a form called RDFa, where "a" is attributes. You could have a similar thing in POI.

alex: How would would we do this?

danbri: We could put it in there, or put a link in via meta links.

<danbri> <meta name="description" content="Alles over Pathé Tuschinski in Amsterdam. Actuele bioscoopagenda, trailers en koop je tickets direct online.

<danbri> " />

alex: It could be something DBpedia returns with a 303?
... A meta tag with a URI or a bunch of URIs.

danbri: Or all the information we need could go in the page, or have links to nerdy computer documents.

jacques_: But it would be using POI as a microformat?

danbri: It could be.

matt: I was just throwing that out there as something we could do, but not that we must.

<danbri> http://graph.danbri.org/Pogo/checker/?url=http://graph.danbri.org/Pogo/checker/testcases/ogp/geo1.cache

ahill: I just want concrete examples, don't care if it is right or wrong.

<danbri> http://graph.danbri.org/Pogo/checker/testcases/ogp/geo1.cache

<danbri> https://github.com/danbri/Pogo/blob/master/checker/testcases/ogp/geo1.cache

<danbri> ... latter has mimetype text/html so you can see it in browser

danbri: OGP is based on RDFa, it can be mapped to RDF and be used in all of the RDF tools.
... Basically using meta they defined their own property types "og:title" "og:type" "og:url" and "og:image", and then you want to 'leverage' the social graph, Facebook traffic, etc, you put that at the top of your doc, then you put a bit of javascript and iframes.
... Whenever someone presses like, the facebook folks scrape this info from the page.
... There are some basic geo things in there, like lat/long and street address, etc.
... So now looking here: http://graph.danbri.org/Pogo/checker/?url=http://graph.danbri.org/Pogo/checker/testcases/ogp/geo1.cache
... You can see we've extracted things from there.
... The good news this stuff is already happening, and it's got a major social network pushing it.

ahill: We have a URI for each of those entities?

danbri: Well here we have some nitpicky things.
... The particular syntax that Facebook chose wasn't about the restaurant itself but about the URI for the restaurant.
... Say there are two pages about the King's Speech, it could be on RottenTomatoes or IMDB. And with OGP, you only have it in one place.
... They wanted it simple, something that you can cut and paste in. They chose not to go the precise expressive route.
... In RDF you figure out what it's talking about by starting at the top of the document, you might have <head about="urn:ietfrestregistry43:"> and then the RDF parser knows to change the subject of each statement it finds.

jacques_: We could consider RDFa representation of an XML format.
... We could use our own XML format and show it in RDFa.

ahill: I can't just take XML and stuff it into this webpage. It's going to be interpreted as a markup tag by the browser.

danbri: Yes, putting XML inside this would not make XML or HTML people happy.
... We'd be better linking XML from it rather than putting it in line.
... Facebook is interested in putting some json, or more within the body of the page.
... Similar to microformats.
... "Welcome to Fidel's fictional restaurant, we're located at <newmarkupforaddress>1234 Mockingbird lane</newmakrupaddress>"

ahill: Is there any sort of established spec for doing what you just described there?

danbri: There are bits and pieces on the microformat wiki. But this probably has more traction.

ahill: Is this not wordy?

danbri: It is meta, so instead of "<meta property='og:countryname'>" you could have "<og:countryname>", but that's not what they chose.

<JonathanJ> Support for RDFa in HTML4 and HTML5 - http://dev.w3.org/html5/rdfa/

<danbri> see http://ogp.spypixel.com/Pogo/checker/

<Carsten> For Ronald: link to OGC navigation paper: http://portal.opengeospatial.org/files/?artifact_id=41727

-> http://graph.danbri.org/Pogo/checker/?url=http://www.rottentomatoes.com/m/matrix/ Matrix movie info

danbri: They've put facebook stuff in and google stuff in.
... the rdf.data-vocabulary.org is from google.

ahill: What's on the left?

danbri: the internal identifiers.
... So the "_:arc*" ones are just assigned internally by the parser.

<JonathanJ> see The Open Graph Protocol - http://ogp.me/

<danbri> <div xmlns:v="http://rdf.data-vocabulary.org/#" typeof="v:Movie"> ...

danbri: There's an ecosystem of laziness.

matt: Presumably there's a checker that you can type your URI into and facebook or google will tell you what it sees.

danbri: If we get a nice clean format you could put stuff in an XML file and then link to it, and then maybe have the conversation with the Facebook and Google guys.
... Or embedding it in KML, or what have you.

ahill: We don't really have RDFa about a movie yet.

danbri: We don't.

ahill: The RDFa we're looking at isn't about the movie per se.

matt: It's not about an instance of the movie.

ahill: In my mind we're adding a little bit of stuff to glue this together from a POI standpoint.
... some glue that in addition to all this information you can find this as a POI.

danbri: The Web folks are getting better about having machine readable information in the Web page.

matt: Adjourn!

cperey: Can we review tomorrow for new attendees?

matt: Sure. I think we should also during this week establish a long-term way to capture our history. We have a blog that we can use. I'd like to put it on the chair, but maybe we can rotate through it.
... Now we're really adjourned!

Summary of Action Items

[NEW] ACTION: andy to start drafting a response to OMA liaison statement to agree with the idea of the liaison and to suggest that they participate in our POI and AR discussion and try to converge their technical specifications with ours [recorded in http://www.w3.org/2011/03/29-poiwg-minutes.html#action02]
[NEW] ACTION: danbri circulate pointer to RDF WG JSON discussion [recorded in http://www.w3.org/2011/03/29-poiwg-minutes.html#action01]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.133 (CVS log)
$Date: 2011/04/05 18:30:14 $