[ gpsdrive ] External program interfacing with gpsdrive

[Hamish]

> > > our hope is, that those mobile devices have enough resources
> > > to run gpsdrive with fullfeatured mapnik rendering and routing,
> > > when those features are finally done...

Hamish:
> > or as the OSM database grows to full coverage. If you want
> > things to scale to larger datasets later on, structure your code
> > decisions so it works on the oldest hardware you can find now....

Guenther:
> that's the reason, why gpsdrive is compilable without the mapnik stuff.
> so it should be possible to run on systems with limited resources with
> a reduced feature set;

or where OSM data is yet lacking.

> but the full functionality is simply not possible.

It seems to me the processing grunt to do routing with a massive
dataset, and OSM quality control to a level which it is useful for
routing will be our biggest hurdle. As a rule I disagree with any
statement that something is "simply not possible". Hard, perhaps,
but never impossible.


> I know that commercial navigation systems do full featured navigation
> in a reasonable time on small systems using less space (about 600MB
> covering europe including POIs on my palm) than osm. if you know how
> they are doing this, please tell us... ;-)

for starters, you cull down the dataset by removing any redundant
vertices while preserving road network topology:
  http://users.ox.ac.uk/~orie1848/tutorial.html
  http://grass.osgeo.org/grass64/manuals/html64_user/v.generalize.html

work with a compressed filesystem; build a fancy network tree so you
only load relevant lines in the realtime processing; coords as floats
not doubles; distribute datasets country by country...


anyway, ways to make things more efficient will come. I don't expect
all to be very slim and fast on the first prototypes, I just ask that
when designing the system keep in mind that goal so that in future
efficency upgrades can be integrated.


glad to see it happening,
Hamish