[Geowanking] Critical Theory

[J. Andrew Rogers]

On Jul 2, 2008, at 4:43 PM, Alan Keown wrote:
> I was struck at the time that managing this data would be a difficult,
> almost incomprehensible, task. In pondering this problem I had the  
> idea that
> location (2D, 3D or 4D as required) provides a unique key for any  
> model
> element in a database. (No two things can occupy the same location  
> if your
> coordinate precision is fine enough.)


The problem is creating a data structure and algorithm that will  
preserve and expose all the underlying spatial relationships that  
unique key represents.  Otherwise, you essentially end up with a hash  
table which is of limited value for most purposes (or we'd already be  
doing that). This turns out to be a hard problem, with thousands of  
pages of published literature spanning decades and no generally usable  
solutions.  You don't want a hash table per se, you want direct  
content-addressability of arbitrary hyper-rectangles in that  
coordinate space, which is (much) trickier under real-world design  
constraints.


> Now that the "Google method" has become spatially enabled maybe they  
> will be
> able to "move from traditional maps to a massive database of spatial
> information like the world has moved from print publications to the  
> digital
> information available on the web" (to quote Landon).


I think you might be misinterpreting this, as Google only supports a  
pseudo-spatial database that is very limited in nature.  See above  
about this being a "hard problem"; solving that problem is  
theoretically equivalent to solving a lot of other important problems  
that would obviate a lot of their existing software infrastructure,  
including the Big Table database architecture.

However, you are correct about the impact, indeed underestimate it, if  
someone built a Google-like infrastructure that generalized to  
arbitrary types and spaces.  The kinds of analytics that could be  
efficiently done are mind-boggling, and it would make the Google of  
today look like how the Google of today makes the old search engines  
of the 1990s look.  But it will take new computer science first, and  
the most interesting work being done in that space is by startup  
ventures rather than an established player.  The circle of life in  
high technology. :-)

Cheers,

J. Andrew Rogers