[OSM-dev] traces, export, download and update

Peter Gervai grinapo at gmail.com
Thu Jan 5 15:38:19 GMT 2012


Hello,

I tried to did the homework and looked pretty hard which made me the
impression that GPS traces are the orphans of OSM. They are hardly
ever mentioned in the developers docs, they are not part of the
architecture schema graphics and only a brief sentence informs us that
it is not "yet" possible to have them in a dump.

Could someone be son kind as to update me about which parts of our
architecture handle these traces, how do we store them? (I can read
code, you can point me to a repository path, too.)

I am familiar with their privacy status, basically they can or cannot
be associated with someone, and they can or cannot provide timestamps.
My guess that this could have been the primary reason dobody seemed to
bother to create a dump of them.

First, there should be a planet.trace.dump, really. It would be so
good and I'd like to see what could I help, that's why I'm asking
about the architecture. Obviosuly they have to be preprocessed to
remove all the private and held back data first.

The other problem related to traces is the API. API can handle the
following cases (apart from uploading):
- retrieve all traces of a bounding box
- retrieve all trace id's of an editor
- getting detail infos on an id
- getting one trace of a given id

The API cannot:
- provide all id's in a bounding box - it would be needed to generate
statistics, for example without retrieving all of the traces
- provide id's of a bounding box which is newer than a given timstamp
or id or whatever is best according to the storage system - see below.
- it would be useful, but not imperative if the detail would contain
statistical data (number of nodes, tracks, labels, etc) on the track

The second missing feature would make it actually not horribly painful
to retrieve an area with traces the second time, since already
retrieved traces wouldn't be needed to be retrieved all AGAIN. If you
worked on an area with some few hundred thousands of points you could
appreciate this.

Without any promise I'm interested in checking whether these could be
implemented. I would appreciate if I wouldn't need to browse _all_ of
the OSM sources to figure out the location. ;-)

Obviosuly if I'm missing a point I would very much appreciated the
hints and advices.

Thanks,
Peter



More information about the dev mailing list