[OSM-dev] Binary OSM; the first pass encoder

[Marcus Wolschon]

Hello Robert,

as far as I understand it the OSM-databases store the location
in long integers anyway. Thus there is no precision to be gained
from using  floats. You only loose storage-space, computation-time
and native functionality on platforms that do not have native floating-
points (like many cellphones, reflashed navigation-devices or
all mobile J2ME java-platforms).

Marcus

2008/11/12, Robert (Jamie) Munro <rjmunro at arjam.net>:
> Floats only allow more precision as the location being considered tends
> towards 0,0 latitude and longitude. For the general case, using an
> integer with the same number of bits of storage will always be more
> precise, particularly at the extremes.
>
> As others have said, 32 bits gives errors of less than 10mm on the
> equator, with even greater accuracy as you move away from it. This is
> around 1000 times more accurate than the data in OSM can be reasonably
> recorded at with standard GPS equipment.
>
> Robert (Jamie) Munro