[OSM-dev] Evaluating the Full-History-Dump

Peter Körner osm-lists at mazdermind.de
Mon Oct 11 22:07:17 BST 2010


Hi

Now that the list of users that accepted the free OdBL is out, it is 
time to do some analysis on the impacts. It is clear that analysis tools 
need to look at the objects' history so the full-experimental-dumps are 
a natural place to start.

But there are some problems with this dump:
  - it is so huge, downloading 18 GBs of data is often a problem
  - it starts with multiple (uncompressed) GBs of <changeset> elements
    that are of small interest for the average user

I'd suggest creating sub-sets of the existing dump that makes it easier 
to play with these things:
  - a dump without the <changeset>-section
  - a dump with only the <node>-section
  - a dump with only the <node>- and <way>-sections

and to get even smaller
  - a dump with only the <node>-section and only with nodes inside a
    certain bbox (eg london or berlin or sth.)

These dumps could be used as test-cases to benchmark tools and compare 
results without the need of a huge 64GB RAM server or 100 MBit Internet 
connection.

Peter



More information about the dev mailing list