[OSM-talk] Potlach problem?

[Richard Fairhurst]

David Earl wrote:

> Do you think it could be that the export takes ages, so the offending
> node(s) weren't in the database when the nodes were exported, but by the
> time the dump reached the segments, someone had uploaded the new set of
> nodes, segments and ways, so the segments and ways appear in the dump
> but not the nodes?
>
> If this is the case, there is the potential for all sorts of
> inconsistencies and integrity problems in the planet file, not just
> missing nodes.

The planet.osm export starts at 2am every day (thanks to Steve for the  
info). If you look at the file timestamps  
(http://planet.openstreetmap.org/?C=M;O=D), it is typically not  
complete until 7am or even later.

So your hypothesis seems sensible. Certainly any dump taking more than  
a second will coincide with _someone_ editing the map, and since we  
don't have transaction-type edits, there'll be at least some  
inconsistency in the data.

That aside, there are, I believe, other known inconsistency issues in  
the database (some, though not all, of which are indeed the result of  
bugs in earlier versions of Potlatch). TomH has been looking into  
these in recent weeks.

Worth checking against next week's planet and seeing if the errors  
reoccur in the same place?

cheers
Richard