[OSM-dev] osm2pgsql for 64-bit IDs

Scott Crosby scott at sacrosby.com
Mon May 23 10:18:37 BST 2011

On Sat, May 21, 2011 at 10:18 AM, Stefan de Konink <stefan at konink.de> wrote:
> Don't forget the binary format (PBF) also was basically twisted around
> to accommodate for negative numbers. Your suggestion makes sense, there
> shouldn't be negative IDs. And additional making the value NOT NULL in
> the database, also gives one bit extra.

PBF has no problems with negative numbers. If there is, that would be
a significant bug  Software using PBF might have problems with
negative ID numbers, but that's not the fault of the format.

If you took an entire planet and negated ID numbers. There would be a
a small hit to filesize because of negative ID's for ways&relations.
About a 500mb/5% increase to the uncompressed file (~60 million
way&relation entities, 2*4 bytes negative number overhead for the
entity ID and, due to delta coding, the first member entity). With
gzip compression, I'd expect <1% filesize hit..

When PBF first came out, osmosis had a few problems when PBF fed it
(-1) as a user ID.


More information about the dev mailing list