The main difference between that approach and what I am doing, is that
I am not using contours: I use the original grid of data points. I am
thinking about calculating local contours on the fly, or use some form
of interpolating the data points. That way I should be able to safe
storage space and it may also be easier to combine it with altitude
data from other sources at some point in the future.

Also, my script adds the (not very spectacular) service of downloading
and verifying the zip files.

I already discovered that my approach eats way to much hard disk
space, so I am changing a few things:

* In stead of storing the coordinates as two double precision floats
(2 * 8 bytes), I will create a bigint index (1 * 8 bytes). Also I
changed the altitude from double (8 bytes) to smallint (2 bytes),
since the SRTM data was already to rounded to the nearest meter.

* In stead of the whole continent I will stick to a smaller part of
Australia for my laptop. I’ll see how far I push my home computer
later.