Points of Interest Working Group Teleconference -- 26 Jan 2011

<trackbot> Date: 26 January 2011

<andy> hi

<Ronald> hi

<scribe> scribe: Matt

<scribe> Chair: Andy

<andy> Meeting Starts Promptly at 9 Roll call Selection of scribe Any objections/corrections to previous meeting minutes http://www.w3.org/2011/01/19-poiwg-minutes Discussion of Location Primitive contribution Is an ID required…Continued from last week. Discussion of Categorization Primitive Face-to-Face Date finalize Announcements AOB

<cperey> +1

Administrative

andy: No objections to last week's minutes.

Review Karl's materials

-> http://lists.w3.org/Archives/Public/public-poiwg/2011Jan/0017.html Karl's email

Karl: I went with overkill for this, figuring it's easier to cut later.

kseiler: Trying to abstract the where from the what
... Tried to decouple description from geospatial location

<cperey> last week we had quite a detailed discussion about ID

kseiler: Within the location primitive there is an identification portion that has IDs, associated IDs, temporal information, update information, how it's used, it's status, and how trustworthy the information is.
... There's a geo-reference portion that can contain a point, which has a type with precision information: high/low, or interpolated.
... Within the point there can be a center point, which has coordinates/height and geodetic system.
... Another portion of the geo-reference can be an address, which has the expected components of an address, e.g. country code and language code, street, but also some things like floor, suite, area, postal code.
... There's also a line portion and 2D polygon, these are collections of points. 3D is left TBD.
... Area, not sure what the best way to do this is, as it could be just a polygon with a center point. Not sure we need a separate thing.
... There's a portion for unknown, it's a POI where we don't know where it is yet.
... Relative, we've done some looking at that elsewhere, we can use that. There's a distance, and bearing, and also an optional reference to one of the other constructs in a POI.
... And a map reference, which lets people not have to provide their own geocoding function. If you have a high fidelity system, you can reference it. e.g. this point is 10% off what OSM has.

andy: Is this overly heavy-weight?
... This is probably the first thing we've had at this level of detail.

<cperey> Alex is saying EXACTLY what I was thinking!

Alex: Thanks for the work Karl. I don't feel like I'm in a position to argue against many of these things based on experience. This seems like a good start.

<cperey> how could you possibly make it more detailed?

<cperey> +q

<jens> more detailed? you could add more indoor directions for instance

kseiler: I'd love to hear if this is supportive of those domain of applications.

<cperey> haha!!

<cperey> good one Alex!

Alex: Knowing where something is, is certainly a good start.
... I don't see anything about orientation. Is anything added assumed to be N/E/W oriented?

cperey: I see at the top that you're trying to isolate the location. Can we say that this information is accurate for this second, or millisecond?
... Someone running say, this could be true for just the instant that the object is there.
... I was working on something the same, but on an object primitive. I like the structure.

<kseiler> +q

cperey: So, I was thinking about how to describe this object with this scheme. An object has an area, a volume, just like what you're saying here.

<andy> +q

Alex: The identification here has this last updated on, etc. You're asking for more information in this area.

kseiler: This description is primarily for locations that don't move. I think it's important to also describe mobile objects that have additional attribution.

cperey: So I think we've agreed that we want something that works for an object as well as a stationary point.
... If you want it for split seconds of time this isn't scalable, we know that. Maybe we could change it to the object primitive and have a timestamp on it and just have it change.

Alex: It doesn't seem to me that we can't accommodate that. Not all of the information is required here, I could imagine it being very lightweight and having just a location, time and a point.

<jens> +1 for alex

<jens> good remark about the optionality on fields

Alex: You could have the location be in flux, while the POI remains the same.
... We could add how long we think it's valid for too.

<kseiler> i think we need to exercise the location primitive via a few XML, JSON examples, from light to heavy

cperey: Then we have the user's position relative to that stationary object.

+1 to exercising it!

Alex: I think that's got less to do with the data model. As long as there's some frame of reference, whether it's WGS-84 or in another one, if I and the object are in the same frame of reference, we're golden.
... I think we'll want orientation, but don't need it right now, as it's implicit in the frame of reference.

kseiler: I ran out of time and bandwidth, but what I'd like to see happen is trying to render out a few use cases with both XML and JSON formatted examples. Lightweight and heavyweight examples. That gives it a better sense of if it's working.
... Do the dog walking through the park.

cperey: Or the coffee cup. We talked about two identical objects that are differentiated by their location at a time.

kseiler: College campuses, or when the coffee shop closes and moves across the street. Big constructs that if there is a problem it's a big problem.

andy: I'm concerned about the direction of the conversation. I think there probably is an object primitive, but I don't think it's the same as this location primitive.

Alex: The identification stuff here is troubling me. What does the owner update? What if the address hasn't changed since 1971, but some new georeferencing came up with new lat/lng

<cperey> muted myself

kseiler: There's an address primitive and a point. The points are getting increasingly refined as GPSes are getting out there. Everyone is standing in their door front with a GPS and getting lat/lng. So, the date changed bit lets you know it's been updated.

Alex: Shouldn't the time go on each piece?

<cperey> what about a machine updating the data?

kseiler: I thought about that, but pulled it out.

<cperey> would machines not be the predominant source of updated data?

Alex: It's not clear to me what protocols people are going to use in general for people to state what information they want.
... We could keep going on and on with heaviness.
... Matt wanted to put timestamps on every field. It sounds valuable, but I don't know the protocol we'll be using for these things to determine what information you get and what is appropriate.

<vinod> +1 to cperey. Periodicity is probably required for a poi in active state. Ex: A hotel poi is active everyday from 10 am till 12 pm. Also the time of existence has to be there Ex: An object in a museum which is active from 8 am till 4 pm and it existed from 1901-1943. Hence I see two more fields in 'Identification' 1.Periodicity for active state 2.Period of existence

<andy> yes

<cperey> and the time stamp which you are referencing

kseiler: We want to be able to turn the dial up and down for how much information you get. Sometimes you want excruciating detail.

<cperey> +q

kseiler: Getting into the area of heavy vs light, will be difficult. Putting things in the spec that give us variability would be good.

<andy> sounds like we have an issue to be opened

Alex: Air on the side of atomiticity. Any element can have any number of attributes attached to it, just they are optional.
... if we have a time atom, we don't really have to care where it goes, it could go up to the POI, or down to the details.
... Instead of us deciding whether it's appropriate for everything to have time info, we just come up with a primitive that allows it to be reused anywhere we want.
... Come up with some basic building blocks that we can tack on to just about everything.
... They don't have to littered throughout the spec everywhere, but just defined once.

cperey: A good way to get an idea of the depth we might want is the Internet of Things. The person I know that would be best for this is Connected Environments. I would take an action item to try to get him involved in this conversation. I'd like to have him come give us a presentation that Pachube uses. They have years of experience in exactly these problems.

<kseiler> good idea

<jens> +1

<cperey> Connected Environments

<andy> +1

<Ronald> +1

andy: Sounds like we have consensus that it's a good idea. Assuming there are no issues with IPR, let's create an action for Christine.

<cperey> absolutely

<scribe> ACTION: cperey to contact Connected Environments to give a presentation. [recorded in http://www.w3.org/2011/01/26-poiwg-minutes.html#action01]

<trackbot> Created ACTION-26 - Contact Connected Environments to give a presentation. [on Christine Perey - due 2011-02-02].

<cperey> I don't know how !

<cperey> Pachube

<cperey> pronouced PatchBay

<cperey> I'lll looking!

http://www.pachube.com/

andy: One thing I noticed is that the location primitive appears to be entirely optional. Can you comment for me?

kseiler: I think we should sweep through this and figure out what we consider optional, etc.

<cperey> Is there anyone who will collaborate with me to develop the Object Primitive which is similar in structure to this?

kseiler: We need to define carefully what is the bare minimum for the location primitive to exist.
... Could be as thin and lean as "unknown".

Alex: Sounds like we're barking up the tree of "at least one of the following".

Keene +

kseiler: I can make a first pass stab at it I can and send it back out for review.

<cperey> yes, deveinitly wiki

<scribe> ACTION: kseiler to try to determine what is optional, required, or one of in location primitive [recorded in http://www.w3.org/2011/01/26-poiwg-minutes.html#action02]

<trackbot> Created ACTION-27 - Try to determine what is optional, required, or one of in location primitive [on Karl Seiler - due 2011-02-02].

andy: I think it'd be good to get into the wiki at this point.

Alex: I'd like to get more clarification about the map reference and where it fits in better.
... If you had a point, a line, and an area, does the map reference apply to all of them equally?

kseiler: Certainly a point. Addresses typically don't exist in maps, they're reverse engineered out of maps.

<jens> +q

kseiler: The address is a linkage to a map.

Alex: Is map reference something inherently useful to show something on a map.

kseiler: It's primarily for navigation and rendering. If you want to know where something is, coordinates will probably tell you. If it's coordinates on a one way street, you need a map reference.

Alex: So if there's something on OSM that indicates a street, then that would be a map reference.
... For instance in navigation there's a way for it to determine that I'm on a street and not in the bushes.

kseiler: Yes, what it's doing is real time perpendicular snap-tos.
... It's a leap forward to "if you know where you are on the map, why keep re-rendering that?"
... What happens in POI listings that are lightly coupled is that they're constantly re-geocoding.
... It's 5-15% error in geocoding. You can squeeze that out by carrying it with you.

Alex: Maybe this could be a form of relative? It's further information like a street address and a 2d point that we care about but it's one of these other constituents like 2d and point.

kseiler: Yes, I like that.

Alex: Find a way to shoe-horn it into relative.

kseiler: I'll work on it.

<scribe> ACTION: kseiler to work on making map-reference as part of relative [recorded in http://www.w3.org/2011/01/26-poiwg-minutes.html#action03]

<trackbot> Created ACTION-28 - Work on making map-reference as part of relative [on Karl Seiler - due 2011-02-02].

jens: Something I noticed missing is that there's no way to specify relative altitude.
... !!
... So you could say that "this is relative to that other location"

Alex: To me, relative is a little bit of a frame of reference question.
... I may want to define points with a lat/lng, but I may want that frame of reference to be relative to another point. So I think Jens is saying we may want the same flexibility in defining something relative as we do absolute.

<cperey> yes, like, relative to something which could be stationary or moving

jens: I think relative might not be relevant to everything but to lat/lng/alt.

Alex: In KHARML, we had a notion of relative to some location. Still use the same fields, but now we interpret it differently.
... One way to accommodate that would be if each of these terms could have an optional attribute that specifies what is really the coordinate system.
... So you might say "Yeah, it's WGS-84, it's meters relative to some other point".

+1 to relative to other points.

kseiler: Is it relative to another point that could be relative, or relative to a point location.

Alex: In general people are going to want to do multi. Basically a scene graph.

<cperey> +1 on scenes/scene graph

<kseiler> ok

<Ronald> +1

Alex: May end up resolving to an unrelative location.

<cperey> Also, I would like to ask what is going to be submitted (if anything) for the meeting on Feb 17

andy: Let's take it up on the mailing list or a call.

Face to Face

andy: The poll says the best time is March 29-31st.
... It will be in the Layar offices in Amsterdam

<kseiler> sorry dates of next f2f?

andy: We're trying to drive down on hotels now. We can arrange for the hotel that's across the Layar offices or look for hotels near the city center.
... I want to confirm that the f2f is March 29-31st at the Layar offices in Amsterdam.

Ronald: Some more information: we have public transport so it is quite easy to get to the city center and back. It's a question of whether you want to be at the meeting early in the morning without the hassle of traveling.

<cperey> sorry that's my battery

Ronald: The hotels have a variety of different rooms with price points.

Alex: Is this a three day meeting?

andy: Yes, thought three days would be better this time. Start later maybe on the first day. Exact hours, I'll send out later in the week.

<kseiler> have to go, adios

Alex: I would say that those of us traveling internationally it's not going to matter if it starts at noon, 1 or 8am. Maybe we hedge towards the idea of the last day being a short day.

andy: My experience the morning in Europe is best to arrive.

cperey: Andy, send me mail on position paper for AR workshop?

andy: Yep.
... I'll send out a note on the f2f soon.

Alex: I don't know about where.

Ronald: It's close to the water, basically just 5-10 minute train ride to central station.

<ahill2> close to Layer sounds fine, we can always travel in the venings

Ronald: We can decide which hotel later this week is my proposal.

<ahill2> in the evenings

Ronald: Can try to get some discount deals for us.

andy: Think we should wrap up. Thanks everyone!

- DRAFT -

Points of Interest Working Group Teleconference

26 Jan 2011

Attendees

Contents

Administrative

Review Karl's materials

Face to Face

Summary of Action Items

Scribe.perl diagnostic output