Some thoughts on Location Tagging
(This whole document is a bit rambly and needs to be tightened up).
At one point, we summarized two key axes for a potential geo microformat:
- Address vs. coordinate systems: I give a location in one of the standard human ways, for instance a street address with a town, state, province, country, etc., or I give lattitude and longitude perhaps with altitude.
- Textual vs. URL standards: An example of textual might be hCard. Textual is data fields expressed in markup that can be displayed, this is more broadly discussed among the location-formats. Our discussion of URL standards in the BOF related almost entirely to URLs that projected onto map coordinates or displayed a location in some way.
If we are going to think about location tagging, we probably need to extend our thinking about how we will use URLs for locations. Also, we may want to be a little less literal in how we interpret location.
Consider neighborhoods in San Francisco or any other city. Normally they have well defined boundaries, well known by real-estate agens. Are the boundaries the same over time or always agreed upon? Humans use the idea of neighborhoods all the time to refer to places that are more or less there. For instance, people more or less know the difference between "Upper Haight" and "The Sunset". They have an idea of what these mean based partially on location and partially on the atmosphere that pervades the location.
Some locations might not "exist" even at this level. Consider Poland between the world wars.
What we might mean by location tagging
Tagging is simply labeling artifacts on the web. In distributed architectures, tags resolve to URLs (see rel="tag") of pages that "define" them. The definition pages might be lists of links or even formal statements of what the tags mean.
Given all of this background (yes, Tantek), location tagging seems to have at least two main interpretations:
Labeling a web artifact as belonging to a location
Oddly enough, I think the most compelling idea is labeling things with the more amorphous locations like neighborhoods. Based on all the talk around Where 2.0 where we had the BOF, location seems to have a special status.
I suggest a very simple extension to reltag that could do this:
<a href="http://servername/tag" rel="tag geo">text</a>
The key point being the use of "geo" as a rel attribute along with "tag". I'll call this geotag. I'll leave it to others to debate whether this should all be done in the rel attribute.
Labeling a precisely defined location
This is simply adding tags to a location. For instance, labeling 16 Fuller Street, Brookline, MA with "first home", "Lily", "Felix", "Angela". You might even add a geotag of "Coolidge Corner".
I would expect that the geo microformat would take care of the precise location in this example.
How it all fits together
Geotag is attractive because it is so simple from a user perspective. No knowing coordinates, no knowing precise locations. The geo microformat, having participated in the BOF, seems oriented toward precise locations like what you would put on a map.
If we look at the evolution of mapping as presented at Where 2.0, however, mapping definitely started in the less precise mode, and in fact continues in that mode today. There is currently a google mash-up of New York burroughs with neighborhoods. Individuals define the neighborhoods (this was at Where as a last minute addition, Nat Torkington would know who these people are). The fact that there is a modern mash-up of this idea legitimizes its currency.
The question is how you marry up precision with the simple amorphous approach in geotag. I see two ways. One, geotag locations in the geo microformat. That essentially attaches a geographic label to the precise location. These can then be aggregated up in the usual tagging way as pioneered by Technorati tags. At that stage, you have a constellation of precisely determined locations defined by the tag, and you can use the constellation of geographic coordinates to define the geographic extent of the tag so that it can be mapped.
Alternatively, you could come up with a definition format for geotag and use that on the page the tag URL points to. I like the first way better because it harnesses social processes.
As for tagging precise locations with words like "first home", that can be easily accomplished by adding reltags to the geo microformat. This works for evaluative tags like "vomitous" and "great". If the tagger decides to make any tag a geotag, then the tag represents a place. One can imagine an aggregation where we have all of the "first homes" listed as places or all of the goods listed as places. This is to be expected and not all that bad (if you think it bad at all). Where geotag gets its value is when place names are socially shared. The key feature of a geotag when you apply it is that you think of it as referring to a place. If we believe all the buzz around Where 2.0, place has importance and deserves a special designator.