Slide 25 of 29
This shows a single object with temporally related “time attributes”. This is hardest to do since every data type has to be temporally-enabled (extension), rather than just adding new temporal data types (addition).
We can envisage multiple interpretation algorithms, so the objects could have several “lines of evidence” for their geometry and attributes even if the original source is a single sequence of rasters.
The <evidence> relationships have to have valid timestamps themselves, and these must somehow be matched 1;1 with the geometry values’ timestamps - or we could just exhaustively search on a time index when we want to match them - or we have defined “event” data-type which is a named timestamp. This holds the semantics of “simultaneity” irrespective of the resolution of the time base (seconds, hours etc., and is even valid for non-numeric time, e.g. “before Tx” or “more than 3 units after Ty”.
Attribute-level time-stamping is done at Munster in the OOGDM project: http://wwwmath.uni-muenster.de/math/inst/info/u/oogdm/
Relationships are the key to implementing everything
