[lug] deconstructing pdf of voted Boulder ballots

[Davide Del Vento]

> There's something like that with GIS imagery as well. I didn't pursue the
> question far but I asked why all of our cells are just 256 pixels square. It
> turns out the third-party tools are much more efficient when working with a
> large number of small images than a few large images.

Especially if you have an embarrassing parallel algorithm (e.g.
pixel-based, not stencil-based) which you can distribute on a even
modest cluster with e.g. MapReduce. But I'm digressing too much from
the topic....