[Data-modeling] Geosearch service (was Re: Modeling uncertainty)

[Tom Morris]

On Thu, Feb 26, 2009 at 2:02 AM, Jonathan W. Lowe <jlowe at giswebsite.com> wrote:

> Suppose you could use points, lines or polygons (or some combination of
> these) to model a location that was "near Oxford".  How would you want
> to mmodel it?  For example, would a circle polygon with Oxford at the
> center represent an area of uncertainty for your purposes?

I way to explicitly encode the degree of uncertainty may be useful in
certain circumstances (e.g. if you've got something like the GPS
dilution of precision to work with), but in the vast majority of the
cases that information is not going to be available.  In those cases,
all that's really needed is a place to store the flag that says this
isn't the actual location, but rather a proxy which is "near," for
some value of near, the actual location.

Perhaps these locations get displayed with a different type of symbol
on maps or perhaps they don't get any special treatment at all, at
least to begin with.  The important thing is to not have to corrupt
the original data to shoehorn it into the system.  How this changes
presentation, search, etc can be worked out over time as more
experience is gained.

Tom