[GeoJSON] Three threads coming together

[Howard Butler]

On Jun 7, 2014, at 9:41 PM, Karl Grossner <karlg at stanford.edu> wrote:

> Jeff,
> 
> I share your frustration; I find that in the context of standards like GeoJSON and now GeoJSON-LD, the cases I want to address are invariably considered 'edge.' In general I think the symmetry between 'where' and 'when' (e.g. Multi-*) is compelling and lends usability. But that view has not won the day, so I retreat and resume work soon from the other direction on a 'PeriodCollection' model (Topotime), with some accounting for "where."
> 
> Overall the introduction of "when" for 'event-like features' is to me a nice step. Don't care for the "circa," but I'm weary of it. Maybe there needs to be a specialization of "when" for trajectory- or path-like features. I don't think MultiInstant alone can get you there in GeoJSON because it breaks the existing pattern; instead it suggests a "where/when" object; which is no longer recognizably GeoJSON.
> 
> I confess that standards are not my game even if models are. :^)

I also support your efforts Jeff. As a casual observer, one thing I think the GeoJSON-LD Time/MultiInstant/Interval effort suffers from is not enough code and full examples. There hasn't been two independent sets of toy code that exercise proposals in any serious way. The bar of general consensus and working code has not been met for the time stuff.

"I like it this way" is true for everyone. "This approach fits my data better" is true for everyone too. One man's edge feature is another woman's core concept. One important aspect that makes GeoJSON successful where others might not be is fully-formed examples. As an implementor, if you simply follow the pattern the examples lay out, you'll get most of the interop you need until your data gets to the edges of the specification. None of the GeoJSON-LD stuff meets that threshold yet.

Finally, is it GeoJSON-LD's scope to be doing for TimeIntervals? Isn't temporal modeling larger in scope than Geo? Hasn't this already been done? Where's there prior art? Why does it belong in GeoJSON-LD instead of in its own thing?

I'll admit to not following along so closely, but can someone take a step back and give the big picture and layout why things have evolved to their current point?

Howard