[talk-au] The trouble with tracing
bluemm1975-osm at yahoo.com
Thu Feb 5 07:48:02 GMT 2009
Nick Hocking <nick.hocking at ...> writes:
> Good tracing is quite helpful in the final definition of a way,
> Bad tracing is really bad - it takes a multiple of the original effort to
> Ok - having resolved that, I believe that the difference between good mapping
> mapping will be absolutely correlated to the number of "one way" errors and
"turn restriction" errors.
> In order to say that a route from Street A to Street B is possible you must
have actually done it yourself.
> It's not good enough to have just passed along a nearby cross street, to
snaffle the street sign name.
> For entitys that need to be routable, you *must* actually travel that route to
verify it's correctness.
> PS - you can't read a "No Right Turn" sign from a satellite and you're not
allowed to use Google Steet Views.
I don't have a GPS yet, so I'm a tracer. I think I've been fairly productive in
my workflow practises).
Recently, I haven't been doing much tracing, preferring to go out there and take
notes down on POI's/surfaces/stop signs/postboxes etc. on a printed out map.
It has been mentioned on the Talk mailing list, but we probably need some kind
of survey_visited=2009-01-15 tag to indicate when someone last went down the
road and verified all the tags/POI's. I am kind of doing it now by using
source:name=survey but that is only for when I have seen the street sign at
intersections. Some way of indicating that I went down this road and
noted/verified all the OSM data would be great.
Also, accurate routing has to be a goal, but it's a lot of effort needed to get
there. Obviously a printable map will come before that, and probably is more
important for OSM to win hearts & minds. Tracing vs GPS tracks from a car doing
60kmh+ will get us nice maps, but not useful routing data. I think of it as
stages or levels of completeness.
More information about the Talk-au