[Data-modeling] Location question

[Richard Newman]

Drat. I had a long reply to this, with my thumb hovering over 'Send',  
when my iPhone received a call and crashed. Oh well. Apologies for the  
lower quality of the retype. The original was much better :)

> I don't think that you can get around the fact that something is
> either part of a set or not part of a set.  If the Northwest and the
> Pacific Northwest and Tom's Rainy States contain different locations,
> they've got to be different things.

I don't dispute that: the issue is that "Northwestern States" denotes  
a concept which itself is associated with different sets of states,  
depending on who is doing the defining — and unlike 'official'  
definitions, nobody is right or wrong.

Either the same name is linked to multiple concrete locations (each of  
which needs a topic ID), which necessitates some kind of horizontal  
association between them, or the name is linked to a single concept  
which needs to be associated with multiple variants.

You would get the same problem if you tried to define a topic for  
"Beautiful women". Locations just happen to be a great example.

> The time-mediation thing is an orthogonal issue to my mind and one
> which definitely needs to be addressed.

Agreed.

> Union vs containment is mostly a matter of semantics, I think.  I
> don't have a problem with the United States "containing" not only the
> 48 contiguous states, but also Alaska, Hawaii, and the Territories,
> unless there's a rule that a Location must be a single closed polygon.
>
> The whole containment thing is currently a mess though because of a
> hack put into the data model/data entry to work around query/client
> limitations.  Because there's no easy way to do recursive containment
> queries, people have been told to enter multiple levels of containment
> in the "contains" property.  This makes it difficult to figure out
> what the true containment hierarchy is (or at least a lot of extra
> work).

I think that union implies an exclusive set of parts, whilst  
containment does not. I agree that the point is kinda moot without  
transitive queries, but with them it would be good to take union down  
to whatever level of granularity seems valid — it's probably fine to  
have countries, states, counties, and even parcels or neighborhoods be  
defined as unions of the next smaller thing, with cities and towns  
etc. being contained by the smallest thing that can contain them  
entirely.

Wow, that was much worse than my first draft. *sigh*

-R