[OSM-talk] New "Highways" view in OSM Inspector
Frederik Ramm
frederik at remote.org
Fri Jan 8 16:42:45 GMT 2010
Hi,
Anthony wrote:
> I'd imagine for some applications we'd want the former (a straight
> road/rail), and for some we'd want the latter (the border of Wyoming).
> Which should be the official definition according to the specs?
Because very few people in OSM have any formal training in geography or
cartography, it is very likely that most mappers - having the standard
slippy map in mind - will assume that a straight line on Earth is a
straight line in the Mercator projection, whereas most programmers will
not hesitate to make the assumption that a line from (100,100) to
(80,80) will pass through (90,90). *Both* of these assumptions are
different, and *both* are not the true shortest surface path (geodetic
or "great circle").
Even if there were an "official definition according to the specs" (ha,
ha), it would be ignored by the majority of people (who simply wouldn't
think that far). Because of that, it is a very pragmatic choice to say
"let's not have distances of more than a few kilometres between nodes"
because that will then reduce the error between the three different
types of "straight line" discussed above to something very small[*].
I think 1km is fine, 5km is acceptable, 10km is stretching the envelope
and anything above that is asking for trouble. OSMI highlights distances
of, I think, 0.3 degrees or more, which translates to roughly 30km at
the equator, or more than 20km in Australia or more than 10km in
northern Canada. Anything highlighted by OSMI is very likely to not show
up on tiles at home maps, and also very likely to not be downloaded when
people download an area for editing in any of the popular editors.
Bye
Frederik
More information about the talk
mailing list