[OSM-dev] history data

Tom Hughes tom at compton.nu
Sun Apr 10 18:03:08 BST 2011


On 10/04/11 17:47, Jochen Topf wrote:

> GET http://api.openstreetmap.org/api/0.6/node/1/history tells me this:
>
> <osm version="0.6" generator="OpenStreetMap server">
>      <node id="1" lat="2.0" lon="2.0" changeset="9257" user="τ12" uid="1298" visible="false" timestamp="2006-05-10T18:27:47Z" version="1"/>
>      <node id="1" lat="50.1249473" lon="14.4557581" changeset="196013" user="tacut" uid="113502" visible="true" timestamp="2009-04-04T10:01:54Z" version="2">
>          <tag k="name" v="Metro C - Kobylisy"/>
>          <tag k="wheelchair" v="yes"/>
>          <tag k="railway" v="subway_entrance"/>
>      </node>
>      <node id="1" lat="2.0" lon="2.0" changeset="524633" user="woodpeck" uid="5164" visible="true" timestamp="2009-04-14T15:42:57Z" version="3"/>
>      <node id="1" lat="2.0" lon="2.0" changeset="1767082" user="Ldp" uid="48796" visible="false" timestamp="2009-07-07T22:44:41Z" version="4"/>
> </osm>
>
> How exactly am I to interpret this? Why are some visible="true" and some
> visible="false"? The visible="false" on the last version tells me the node
> is deleted. But what happened before? Does the visible="false" on the first
> version mean it was deleted before it was re-created as version 2? For a
> deleted node, does the timestamp refer to the creation or the deletion?

Here goes...

The real node 1 should be somewhere in the Regents Park area of London 
but it has a history of being accidentally changed by badly written 
imports and bots which just make up IDs starting at 1 and things.

It looks like in May 2006 it was deleted - it was already at a silly 
location so had probably already been the victim of a rogue client. We 
have no history before that online because history was wiped when we 
went to api 0.5 in October 2007.

It was then resurrected by somebody uploading a valid node 1 in April 
2009 (just a few days before api 0.6 went live with version checks to 
make these accidental overwrites harder).

Frederik obviously noticed this and reverted that changeset, moving the 
object back to it's previous position again but not actually marking it 
as deleted. Eventually Lennard came around and deleted it again.

If you look at changeset 196013 you'll see it is just an upload of nodes 
1 through 17 so a classic bad import. Those 17 nodes are then moved back 
again in Frederik's changeset.

Tom

-- 
Tom Hughes (tom at compton.nu)
http://compton.nu/



More information about the dev mailing list