[OSM-talk] [OSM-dev] gpx upload sanity check please

[Stefan Baebler]

> >> 1,000km even logged at 1 point per second and travelling at walking pace
> >> should only produce about about 1 million points.
> >> Is the 1,000km figure at all accurate?
> > It is:  1pt/s * 1000km / 4km/h = 900000pt
> I agree: 1pt/s * 1000km * 1hr/4km * 3600s/hr = 900000pt
> Just in case anybody else had a brain cramp....
or google's calculator
http://www.google.com/search?q=1000km%2F4km%2Fh+in+s

However i somehow doubt the 1000km trace was made by a 10 day nonstop walk.

>From http://www.openstreetmap.org/user/user_1400/traces/ it seems that
these are many repeated journeys with slight variations, adding up to
staggering 7 million gps readings

Driving 1000km or more at higher than walking average speed 52 times
seems possible (not sure of mapping usability though).
http://www.google.com/search?q=52+*+1000km%2F30km%2Fh+in+s
(52 * (1 000 km)) / (30 km/h) = 6 240 000 seconds

http://wiki.openstreetmap.org/index.php/Database_schema#gps_points
I didn't check if tracks overlap, but x,y and timestamp should
probably be globally unique, not just per gpx, so making a unique
constraint with x,y and timestamp could make sense (the chances for 2
GPS units recording the same position at the same time is practically
zero)...but I'm not sure what this means in terms of performance.
Maybe this new constraint could simply replace the existing primary
key.

For improved reading performance i guess there must already be a
_separate_ index on each x and y (and possibly gpx_id).

Stefan