[OSM-newbies] Starting with JOSM and OpenStreetMap in general

Rick Collins gnuarm.2006 at arius.com
Mon Oct 8 21:55:17 BST 2007


At 02:34 PM 10/8/2007, you wrote:
>graham wrote:
>
> > Rick Collins wrote:
> >> I think this is what I don't understand.  Why do I need to "create"
> >> nodes when I already have all these points collected from my
> >> GPS?  Clearly I don't want to have to duplicate all of these
> >> points?  Or are you saying that I need to trim them down to a minimum
> >> that still follows the course of the road accurately?
> >
> > Yes, that's it.
> > There are (at least) three reasons for this:
>
>1. You may have a lot of gps points along a straight piece of road, 
>which could be represented with just two nodes. Using the gps points 
>as nodes would multiply the amount of storage needed hugely, as well 
>as complicating things for any software which manipulates the nodes 
>(renderers, route finders etc)
>[...]

Graham, I don't really agree that any of these reasons justify 
*requiring* all points to be manually entered.  Yes, I found a few 
points that needed to be cleaned up and I did that in another program 
that superimposes my tracks on USGS photos and maps.  But a track may 
have thousands of points.  I have no interest in manually translating 
this into OSM data.


>The one scenario when it's really useful to automatically convert
>from a GPS track to an OpenStreetMap way is on long, winding rural
>roads. GPS accuracy here is generally pretty good (no concrete
>canyons), and tracing 100 miles of bends is not fun.

I agree that most collected points are good enough.  In fact, I 
assumed that this project would harness the power of the many 
computers available on the Internet and use an algorithm that would 
accept multiple measurements of the same roads and maintain an 
average or possibly just toss out stray points.  I rather imagined 
that it would be a database of a huge amount of collected data with 
minimal human input.  Then computing capability would provide an 
automatically culled, much smaller data set for users and routing 
software to make use of.

Just how large is the current data set?  Multiple terabytes?  If not, 
I don't see this project ever really stressing the rapidly increasing 
storage capability of a small number of PCs.


>With Potlatch, you can upload a track to OSM, then automatically
>convert it to a way. Potlatch "simplifies" the way to remove surplus
>points while retaining the curves. You will, of course, still have to
>tidy it up a little yourself - splitting the way at the right points,
>making junctions with other ways, etc. - before saving to the database.

As I said in my post, I don't use Flash Player since I am on dialup 
and the videos it works with far exceed the capacity of my link.  In 
order to avoid web pages that gratuitously devour my small bandwidth, 
I just don't have Flash Player installed at all.  So I can't use 
Potlatch.  It just seems much more efficient to use a program that 
initially processes the data offline.


>Details here: http://wiki.openstreetmap.org/index.php/Potlatch/
>Primer#Working_with_GPS_tracks
>
>A quick Google suggests that a similar feature was experimentally
>available in JOSM at one point, but any more than that I don't know.

Well, it's too bad it was not retained.  To be honest, I am not very 
fond of working with Java applets at all.  They never seem to work 
very well.  I understand the advantage of being portable without 
having to maintain multiple versions for multiple OS and CPUs.  But 
any time you try to design to a lowest common denominator, it does 
not work well on any machine.  Regardless, JOSM is workable, but it 
doesn't do what I feel is needed for this effort.

Does anyone know of a way to automatically convert a GPX file to the 
format required by OSM?  OSM seemed to allow me to upload GPX 
files.  If I can't use Potlatch, is there any use for them?






More information about the newbies mailing list