Precision and accuracy in features

[Rob Warren]

I have been dealing with a lot of historical locations (some of which no longer exist) and trying to express them using linked open data.

How are people dealing with the problem of precision and accuracy in publishing data? A number of data sets are using points or boxes to identify locations and this is "good enough" for display but a bit problematic for retrieval.

I'd like to be able to document the error generated when tracing from hand-drawn maps or when converting positions from older systems such as the British Trench Map coordinates.

So far, the best I can come up with is to have a second geometry that maps out the error region around the original location. Thus for a given gn:Point, you would map out an error box centered on it that would model alternate potential locations.

Has anyone else been dealing with this problem and would you care to swap ideas?

Rob