W3C home > Mailing lists > Public > public-poiwg@w3.org > November 2010

Re: resend: definition of a 'thing'

From: Alex Hill <ahill@gatech.edu>
Date: Wed, 17 Nov 2010 10:15:28 -0500
To: "public-poiwg@w3.org W3C" <public-poiwg@w3.org>
Message-Id: <1B3B8E52-28A9-4E45-A868-58F9A13DEEEB@gatech.edu>
As per our discussions today, we seem to be in agreement that specifically including NFT, RFID or other sensors into the spec is not in scope.
But we also seem to agree that non-spatially or non-temporally static POI's definitely need to be covered by the spec.
We main not agree exactly how to describe now to track this point, but we need to give some form of URI or additional attributes that would allow it.
I think that a URI that describes how the phone would go about identifying and tracking an object is reasonable.
I think a URI that points to a service that will report the location of that object (i.e. milk truck) is reasonable.

There was also a suggestion from Gary that a POI without any sort of modifiers we considered static and permanent.
Christine and I objected to the suggestion that the "majority" of POI's will be static and permanent in the near future, but I think we agree that the data format need not be accused of dictating such assumptions.
From the phone meeting IRC:
"<Alex_>Just because the data format with no modifiers means a static and permanent POI doesn't mean an endorsement that the majority of POI's are of this type ...
 ... those temporally short or dynamically tracked POI's will have more attributes in their respective databases but if they are no longer here in time or space the query should not return them (unless you want it to)."

On Nov 10, 2010, at 5:03 PM, Rob Manson wrote:

> Hi Mike, there was quite a bit of discussion early on about including
> non-geo sensors...but the terms of reference for the working group were
> specifically scoped down to exclude other sensor based detection like
> NFT and audio detection, etc.  
> Otherwise we're back to talking about Patterns and not Points again,
> which would suit me fine btw 8)
> But I think Alex's ending question is right and that's why I suggested
> that in the levels of abstraction we also clearly define what is in and
> out of scope.
> Gary, I think your email provides a great summary of the key terms.
> Based on this it seems pretty clear that in the context of #poiwg,
> Location is the key attribute used for POI discovery, search and
> presentation (e.g. placement).  And that POIs are really just a way of
> relating arbitrary Information to Locations so they can be found and
> re-presented.
> Then the problem is simply a matter of defining the approach for
> creating these relationships (e.g. a semantic discussion - perhaps based
> on triplet style encoding).
> And the other simple problem is what metadata is consistently available
> for re-presenting these POIs.  And here the existing dc: type metadata
> models seem like a great starting point.
> This way there is a clear abstraction between the real world and the
> digital world.  
>        The real world contains things that can be dropped on your foot
>        and that occupy real physical and temporal spaces.  
>        The digital world contains a set of POIs that link arbitrary
>        information (content and meaning) to Locations (or probably more
>        specifically centroids) along with a consistently re-presentable
>        set of metadata (including optional 2D, 3D and audio
>        representations).
> For me the richest source of information that I want to be able to
> relate to Locations through POIs are web pages.  I'd love to be able to
> add one or more rel type links (for example) to a page to make it
> available as a POI or collection of POIs.  This would allow systems
> (e.g. AR browsers or search engines, etc.) to index pages by Location
> and then use the extracted metadata for re-presentation.  This would be
> a truly w3c style POI standard that really embraces the web instead of
> just using it for it's HTTP transport layer.
> NOTE: Thomas' point about other transports is really valid and important
> too though.
> PS: Gary, one key thing that is missing from your email is something to
> do with control over the re-presentation of the POIs too.  
> e.g. orientation, scale, etc.
> roBman 
> On Wed, 2010-11-10 at 10:22 -0500, Alex Hill wrote:
>> It seems to me that we could divorce the definition of a "location"
>> from the sort of technology that supports it.
>> In the future, everything will be tracked by some means (I am already
>> being tracked at a coarse level; credit card receipts, GPS, etc.).
>> Whether it is the device recognizing the item, partially recognizing
>> the item and using a database to improve accuracy, or a direct signal
>> indicating its location, it is still a "thing" that needs some sort of
>> URI.
>> And since it is a part of the physical world, it can be the "location"
>> for a POI.
>> This is why I would lean towards incorporating "thing" into "location"
>> if possible.
>> Here is an example use-case:
>> My device sees a rock. I decide it is a POI and attach some data to
>> it. The "world" or "cyberspace" may know nothing about it, yet.
>> Yes, it may conveniently be big and difficult to move. Therefore it's
>> "location" description is quite simple.
>> However, if it is small enough to be moved, its "location" description
>> gets more complex.
>> If a sensor has been affixed, then tracking its location becomes
>> straightforward.
>> However, if only the information the URI includes is what my device
>> saw then the "location" of this item is intermittent, perhaps
>> "unknown", for periods of time.
>> There will certainly be services that use a combination of credit card
>> receipts, last GPS location and face recognition to give a device my
>> "location".
>> Seems we need to decide early on if we want to capture this in the POI
>> standard.
>> On Nov 10, 2010, at 9:51 AM, Mike Liebhold wrote:
>>> Here's draft definition of 'thing'
>>> A ' thing is a physical object with no fixed location that may have
>>> a POI digital information attached that cannot be automatically
>>> detected by geopositioning sensors. The POI,therefore will only be
>>> detectable by a default visual ( or acoustic?) search, invoked by a
>>> client.
>>> ***
>>> Question: Does this require a default registry of visual or acoustic
>>> search services that can decode the visual ( or acoustic?) pattern
>>> to point the client to the appropriate URI that explains the POI
>>> format?
>> Alex Hill Ph.D.
>> Postdoctoral Fellow
>> Augmented Environments Laboratory
>> Georgia Institute of Technology
>> http://www.augmentedenvironments.org/lab

Alex Hill Ph.D.
Postdoctoral Fellow
Augmented Environments Laboratory
Georgia Institute of Technology

Received on Wednesday, 17 November 2010 15:16:04 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 16:48:26 UTC