[OSM-talk] conclusions from my postgres / postgis experiments

Richard Fairhurst richard at systemeD.net
Thu Nov 23 14:02:48 GMT 2006


raphael Jacquot <sxpert at sxpert.org> wrote:

> all those people use a standard format to 'share' their data, GDF
> (Geographic Data File), which was specified by the CEN technical
> committee 278 Road Transport and Traffic Telematics which defines such
> a structure.

Sure, but they don't use that standard internally. When you buy a  
satnav box for £199/300eur, it doesn't have a full GDF-format dump  
inside it. The TeleAtlas data, shipped as GDF, has been processed by  
TomTom/whoever into a lean, fast database. On the URL you cite:

   "A GDF database (in the exchange format) will never be used as
    such. Users will first transform it into their system, which
    could be a GIS (Geographic Information System), car navigation
    system, or routing algorithm. Therefore you cannot talk about a
    GDF application, but rather an application that uses GDF.

   "GDF is not a CD standard for car navigation systems. GDF is
    used and converted onto the CD-ROM in the internal format of
    the navigation system."

The geodata actually exists in four separate formats:

   1. Original Ordnance Survey MasterMap data (for UK)
    -> aggregated as part of
   2. TeleAtlas master database
    -> distributed to
   3. TomTom plc as GDF file
    -> interpreted into
   4. TomTom satnav unit internal format

The master OSM database is closest to 1, maybe 2.

If I read your Grenoble example correctly, you were saying that OSM  
should use a particular format (superways) because it would be more  
efficient for route-planning apps, like satnavs (4). I think the  
existence of GDF, rather, proves that the storage data format, 1 (=OSM  
db) is different to the client data format, 4 (=routeplanner internal  
format), and that the ingenious halfway house of GDF, 3, has been  
devised to ease the transfer of one into another. Maybe one day we'll  
have planetosm_to_gdf.pl?

cheers
Richard





More information about the talk mailing list