- From: Matt Womer <mdw@w3.org>
- Date: Wed, 27 Apr 2011 09:03:43 -0400
- To: public-poiwg W3C <public-poiwg@w3.org>
Hi all,
The minutes from last weeks minutes are available here:
http://www.w3.org/2011/04/20-poiwg-minutes.html
And as text below.
The main topic was ISSUE-3, in the form of "Are POIs tangible?"
-Matt
[1]W3C
[1] http://www.w3.org/
- DRAFT -
Points of Interest Working Group Teleconference
20 Apr 2011
[2]Agenda
[2] http://www.w3.org/mid/B7B2E88280255C40A477C45E46CF6C25241DDCECC2@usrtmbx01.corpusers.net
See also: [3]IRC log
[3] http://www.w3.org/2011/04/20-poiwg-irc
Attendees
Present
Matt, +3539149aaaa, vinod, Alex, Carl_Reed, Andy,
+1.312.894.aabb, Karl, Fons, +1.617.764.aacc, Raj,
+1.617.848.aadd, Luca, cperey, IanPouncey
Regrets
danbri, Ronald, Carsten, Gary, Jens
Chair
Andy
Scribe
Matt
Contents
* [4]Topics
1. [5]New call time poll
2. [6]Are POIs Tangible?
* [7]Summary of Action Items
_________________________________________________________
<trackbot> Date: 20 April 2011
<danbri> (I have my standing class w/ the Semweb CG call now, which
I need to join, but i'll read along here)
<danbri> so regrets from Dan B
<scribe> Agenda:
[8]http://lists.w3.org/Archives/Member/member-poiwg/2011Apr/0029
[8] http://lists.w3.org/Archives/Member/member-poiwg/2011Apr/0029
<Luca> +Bjorn_Bringert is Luca
<Luca> +Luca
<scribe> scribe: Matt
New call time poll
-> [9]http://www.w3.org/2002/09/wbs/45386/POIWG-new-call-time/ Poll
[9] http://www.w3.org/2002/09/wbs/45386/POIWG-new-call-time/
Andy: I'd like to wrap this up, so please answer.
Are POIs Tangible?
ISSUE-3?
<trackbot> ISSUE-3 -- Core POI spec should provide a way to indicate
families of POI, such as commercial brands (eg. starbucks) -- raised
<trackbot> [10]http://www.w3.org/2010/POI/track/issues/3
[10] http://www.w3.org/2010/POI/track/issues/3
Andy: This started as a question about organizations rather than
tangibility, but that's where the thread led.
... The list consensus seems to show that for instance, Starbucks
could be a presence, but it's not necessarily the concept.
<ahill> +q
karls: POI names change all the time. You still need a metadata
association. Does that concept of a chain and it's descendent
children exist in our spec?
Carl: I don't disagree that metadata is needed, but it opens a huge
can of worms and increases the complexity.
... There was a meeting on gazetters last week, Raj was there.
<cperey> +1 on getting more use cases and inputs from people who are
working in this domain
<cperey> a URL to this workshop Raj attended?
Raj: There are many people working on large historical gazetters.
They talked about the need for a concept with no geography that
covered everyones idea of a place. A place that may change over
time, or have a different geography than those of the
administrators.
... For example: Hampshire, which had no physical boundaries, but
also a physical boundary, and a place that everyone knew as
Hampshire.
<ahill> +1 carl
Raj: If you had metadata for this you could do something like link
to Dun and Bradstreet.
<cperey> linking into third party data sets is practical, or what
aspect is not practical?
karls: I'm looking to keep this simple. Example, Starbucks, there
could be a POI for Starbucks that then points to Starbucks the
corporate entity that could then be pointed to by all of the
Starbucks.
... It's similar to a POI, but doesn't always have a location.
<ahill> what is the relationship between Categories and Corporate
Entities?
Raj: We're not going to be able to do that as well as the people who
for instance collect taxes. I do see the usefulness of the unique ID
with real good information, but we're not going to be able to do it
as well as an authoritative source.
karls: D&B has the concept of a chain, they can point to other POIs.
cperey: What isn't practical about doing what is just suggested?
Linking to a third party data set?
matt: I don't think it was that we weren't going to link just that
we wouldn't define the meaning.
karls: Linking is a core requirement.
... I think we're talking about whether POIs have to have a
location. If they don't, they can represent concepts. It seems that
not requiring a location was opening a can of worms.
... For our intents and purposes then it must have one location.
... Anything else we can point to.
<fons> +1 to POIs having one locatio
matt: I don't think I agree that they must have a locationn. Can
someone expand on that?
<Bertine> +1 to needing location of some form
karls: I think we decided that at the first f2f, it's how we have
been documenting it.
<Bertine> (else we'll just get into linked data and semantics which
other people are solving
ahill: We've talked about having one location, but also having
multiple descriptions of that location.
... The distinction here is whether something is an entity that
exists in the physical world, it might be a place that we only know
it's location relative to something else.
<cperey> Alex said that we have agreed an entity has to have a
location, but the location can be unknown
ahill: So it is something that is physically located.
<Bertine> Disagreeing: must be a known location, even if it's not
fixed
cperey: I think it's important that it have a location, but that it
can be unknown.
<Bertine> (or relative)
<rsingh2> +q
karls: So, if we all agree with that, we can have something like a
corporate entity that we can then chain to that which doesn't have a
place.
<cperey> is Bertine on the call and want to input?
Carl: Christine mentioned relative -- GeoPRIV at IETF has worked on
issues related to relative location and uncertainty.
... Those are two important topics dealing with points.
<Bertine> All I would like to point out is that there is no real
//meaning/ to an unknown location
<scribe> ACTION: Carl to send WG pointers to Geopriv documents that
deal with location [recorded in
[11]http://www.w3.org/2011/04/20-poiwg-minutes.html#action01]
<trackbot> Created ACTION-64 - Send WG pointers to Geopriv documents
that deal with location [on Carl Reed - due 2011-04-27].
<fons> +1 to Bertine
karls: I think the motivation for this was that we know a place
could have a location, we just haven't resolved it yet. We still
want to be able to hold the entity.
<rsingh2> Ordnance Survey (UK) is publishing a place name database
of the UK. They will come out with a new version in a couple years
and see an "intangible POI" as a possible way of helping with
versioning -- helping people who link to their POIs link to the new
ones
<Bertine> I'm not really sure why a half-written document would be
required in the spec. We don't have half-written HTML :p
ahill: What is the relationship between categories and corporate
entities?
karls: I tried to express this in an email that put some meat on the
bones of the category primitive.
... You can build the same functionality as categories by linking. I
struggle with categories as every time we deal with the data, we hit
different schemes and then have to map to a standardize scheme, etc.
It's all highly subjective and interpretive. I think we're on the
edge of search meta data, keywords, etc. I often think maybe we
could give it up.
ahill: Can we come up with a resolution about the relationship
between categories and "authoritative sources"?
... Is it possible that one is less authoritative, and have less
value?
karls: I would propose that the category primitive is optional and
can be flat or hierarchical.
... That it can point to authoritative source URIs, it could link
out.
<Bertine> +1 to linking
ahill: It's possible that in the category tag, there could be
similar URIs to other authorities, those that keep track of
categories. Is that right?
<vinod> Hi, Michael and myself were looking at categories and
Michael suggested using the existing 'Category knowledge'
<vinod> [12]http://www.w3.org/2004/02/skos/
[12] http://www.w3.org/2004/02/skos/
karls: Everyone has their own proprietary category system that
facilitates finding things. Then there are authoritative sources
from emerging business standards. Those are bounded in scope to
business categories, not things like parks.
... You need something to bridge between the two.
<vinod> Even DBpedia categories(i.e Wikiepdia Categories) seemed
interesting
karls: When you talk about chains, you are talking about membership
in an association. I am not sure they are the same thing.
<vinod> Thats OK. :)
<Bertine> If you link to an entry in an authorative database would
not that database have a catagory for the entity?
matt: I could see us having the category be collapsed into
relationships, e.g. "<relationship type='is-a'
link='[13]http://starbucks...'/>"
... Anyone want to write up some examples.
[13] http://starbucks...'/%3E
ahill: There has been some discussion about having relationships
stay specific to some of the location related aspects of POIs, while
you suggested that the relationship primitive could be used for
everything.
<Bertine> The two should absolutely be seperated. They're two
completely different things
ahill: So you are suggesting that POIs could be conceptual things.
Are we in agreement on how far down the road we want to go with
relationships? Concepts? Groupings?
<Bertine> Having something be positioned as a child is completely
different from being a child of business chain.
Raj: The Atom specs have a great thing on picking authoritative
categories.
ahill: Matt was suggesting that the relationship primitive be used
to describe something bigger than just physical relationships.
... I am a worried about using the relationship primitive to manage
these things.
karls: I go back to the theater district example, very vague area.
<Bertine> +1
karls: How do you link to it? What's the POI for the theater
district, polygonal? Is that legal?
ahill: I don't see that as problematic.
... I was thinking more of using the relationship primitive to
establish say a relationship between a McDonalds and the McDonalds
corporation.
Carl: Gazetters argue this too. London, is it a point? They use the
boundary of London, a polygon, to represent the "point".
<rsingh2> information on atom:category is at
[14]http://www.atomenabled.org/developers/syndication/#category and
[15]http://www.ietf.org/rfc/rfc4287.txt (clause 4.2.2)
[14] http://www.atomenabled.org/developers/syndication/#category
[15] http://www.ietf.org/rfc/rfc4287.txt
<rsingh2> +q
ahill: Some authority has said "this is London's boundaries" -- how
do you deal with that authority?
<Bertine> Ryan air can claim to be anywhere and there is not much we
can do about it
ahill: POIs may claim they are in London, but I can see London
having a polygonal boundary that doesn't include that POI. What do
we do in that case?
karls: To facilitate spatial search, "is this x y in London?" is
usually based on the official boundary of London. Then there's
administrative hierarchies: country, city, neighborhoods, etc.
There's a hierarchy that varies from country to country.
rsingh2: There is the physical relationship, touching, etc. The
gazetteer workshop talked about this too.
karls: It's not in our charter to define all of that, we should hook
in, facilitate where POIs are -- topological search, geographical,
or administrative based searches.
<Bertine> Locations specifications could allow links i.e. link to a
remotely hosted (official) polygon of an area
rsingh2: In Atom's category element you have the name of the
category, then a pointer to the scheme used for categorization.
... So you can point to an authority or have free-form tagging.
... So you could pull from say dbpedia.
... I think this is different than talking about relationships
between POIs.
karls: It's similar. Categories are a description of a type of
thing, which you could do with relationships, but relationships are
used for representing membership, etc. I think it's valuable to keep
them separate.
<vinod> +1
ahill: Is there value in extending a relationship to a category or
not?
rsingh2: Not in 2011....
<cperey> this was a good discussion
ahill: We may be close to something here. We've come back around to
the initial topic in some sense.
<vinod> if we are depending on db/wikipedia categories, we would be
letting users create POIs which don't have location attribute .
<vinod> !!
ahill: It sounds like we have a consensus that we're not going to
represent corporate relationships. That we might have metadata for
that. And I would argue that we would restrict the relationship
primitive to things like "the theater district".
<cperey> relationship primtive can encapsulate categories
<cperey> and concepts are dealt with via categories
PROPOSED RESOLUTION: Points of Interests have a location in the
physical world, they don't have to be entities, but they have a
physical location.
<fons> +1
<vinod> Well
<Bertine> +1
<karls> +1
<Bertine> (well, +2 technically :p)
<Luca> +1
vinod: If you're relying on a real world category like DBpedia, then
we would use those to create POIs without a location.
<cperey> are we going to say that the physical location can be
"empty"?
<cperey> location=unknown is permitted
matt: I think we were going to have a location primitive anyway, but
with an unknown location.
vinod: I think we should discuss more.
<Bertine> -2 for the unknown location - really doesn't serve any
purpose. Placeholders shouldn't be in there
rsingh2: And in the categories we would have things like "this is a
coffee shop".
vinod: So can something have no location in a POI?
rsingh2: Yes, but you try not to. But the categories are not the
POIs.
<Bertine> Why would we be creating POI's for all wikipedia pages?
How does 'Apple' benefit having a POI?
<fons> -1 for location=unknown is permitted
<karls> we are saying a POI MUST have a location, it could be
temporarily unknown
vinod: Categories will have no location in say Wikipedia. If we
create categories based on real world things, we would have
categories that don't have a location.
... There are categories, say I create a POI and attach a category
to them, but that category won't have a location.
karls: We're saying that a POI must have a location, it can be
unknown, but it must have one. Categories do not have a location.
rsingh2: A category could be "has handicapped access" or "made of
brick".
ahill: I don't think our plan is to have the POI spec be able to
describe everything in wikipedia.
->
[16]http://www.w3.org/2010/POI/wiki/Core/Draft#Location_Attribution_
Details Location Primitive
[16] http://www.w3.org/2010/POI/wiki/Core/Draft#Location_Attribution_Details
->
[17]http://www.w3.org/2010/POI/wiki/Core/Draft#categorization_primit
ive Category primitive
[17] http://www.w3.org/2010/POI/wiki/Core/Draft#categorization_primitive
<cperey> +1
matt: Has everyone seen what has been written?
... Can everyone read those and send feedback to the mailing list?
... I would like to have a good draft for next week of the FPWD.
<karls> bye bye
matt: I think rather than digging into the next agenda item that
we'll just adjourn now and talk next week.
... The call time for next week will remain the same, unless it is
announced on the mailing list.
... Please fill in the poll at:
[18]http://www.w3.org/2002/09/wbs/45386/POIWG-new-call-time/
[18] http://www.w3.org/2002/09/wbs/45386/POIWG-new-call-time/
Summary of Action Items
[NEW] ACTION: Carl to send WG pointers to Geopriv documents that
deal with location [recorded in
[19]http://www.w3.org/2011/04/20-poiwg-minutes.html#action01]
[End of minutes]
_________________________________________________________
Minutes formatted by David Booth's [20]scribe.perl version 1.135
([21]CVS log)
$Date: 2011/04/20 13:56:04 $
_________________________________________________________
[20] http://dev.w3.org/cvsweb/~checkout~/2002/scribe/scribedoc.htm
[21] http://dev.w3.org/cvsweb/2002/scribe/
Scribe.perl diagnostic output
[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.135 of Date: 2009/03/02 03:52:20
Check for newer version at [22]http://dev.w3.org/cvsweb/~checkout~/2002
/scribe/
[22] http://dev.w3.org/cvsweb/~checkout~/2002/scribe/
Guessing input format: RRSAgent_Text_Format (score 1.00)
Succeeded: s/Dunn/Dun/
Succeeded: s/locatio/location/
Succeeded: s/ahill2/ahill/g
Found Scribe: Matt
Inferring ScribeNick: matt
Default Present: Matt, +3539149aaaa, vinod, Alex, Carl_Reed, Andy, +1.3
12.894.aabb, Karl, Fons, +1.617.764.aacc, Raj, +1.617.848.aadd, Luca, c
perey, IanPouncey
Present: Matt +3539149aaaa vinod Alex Carl_Reed Andy +1.312.894.aabb Ka
rl Fons +1.617.764.aacc Raj +1.617.848.aadd Luca cperey IanPouncey
Regrets: danbri Ronald Carsten Gary Jens
Agenda: [23]http://www.w3.org/mid/B7B2E88280255C40A477C45E46CF6C25241DD
CECC2@usrtmbx01.corpusers.net
Found Date: 20 Apr 2011
Guessing minutes URL: [24]http://www.w3.org/2011/04/20-poiwg-minutes.ht
ml
People with action items: carl
[23] http://www.w3.org/mid/B7B2E88280255C40A477C45E46CF6C25241DDCECC2@usrtmbx01.corpusers.net
[24] http://www.w3.org/2011/04/20-poiwg-minutes.html
WARNING: Input appears to use implicit continuation lines.
You may need the "-implicitContinuations" option.
End of [25]scribe.perl diagnostic output]
[25] http://dev.w3.org/cvsweb/~checkout~/2002/scribe/scribedoc.htm
Received on Wednesday, 27 April 2011 13:03:53 UTC