[OSM-talk] Redaction requests - the information I need

[Paul Norman]

> From: data at osmfoundation.org [mailto:data at osmfoundation.org] On Behalf
> Sent: Wednesday, November 14, 2012 12:43 AM
> To: Paul Norman
> Cc: Talk Openstreetmap; data at osmfoundation.org
> Subject: Re: [OSM-talk] Redaction requests - the information I need
> 
> On Tue, Nov 13, 2012 at 11:44 AM, Paul Norman <penorman at mac.com> wrote:
> > 1. Deletes data as required using changesets. In many cases this
> > essentially the same as reverting the changeset so if someone has
> > already reverted it nothing is done here.
> >
> > 2. Uses the redaction API call to hide old versions of the objects
> > that cannot be shown.
> 
> Does it mean that 'bad data' is removed from the database but not from
> the planet dumps ? and what about all db replications around the world
> if they keep the history of elements based on timely diffs ?

Planet dumps only include visible objects so just the deletions will cause
the objects to not be included.

Full history dumps only show unredacted objects so they're okay.

A service mirroring the latest data (e.g. a local planetfile,
pgsnapshot+jxapi or overpass) is also okay, assuming the replication process
works and there are no errors.

osm2pgsql should also be okay although there are bugs that can occur in rare
cases with conversions between multipolygons and normal ways. I don't
believe the redaction bot can trigger these cases.

There is not sufficient information in the replication diffs to keep a local
apidb in sync with the osm.org API. The local mirror will continue to return
history for redacted objects.

I can't give any definite answers about other DB schemas. Generally anything
where you could drop the database and reimport the latest data and end up
with an equivalent result to if you had used replication will be fine.

My guess would also be that services like Pascal's http://hdyc.neis-one.org
are also fine as it's not really presenting OSM data.

If this is a concern about the local mirror on openstreetmap.fr I don't
believe it's currently an issue if it only has France in the DB.
http://yosmhm.neis-one.org/?u=pnorman%20redaction%20revert doesn't show
anything in France and I think the nearby changesets in Germany were all
pre-license change.

If the server does store history there is the possibility of an issue. The
proper way to fix this would be to add information about the redaction API
calls to the replication diffs in a backwards-compatible way. An easier
solution only applicable to a caching server might be to discard the history
of an object whenever pnorman redaction revert touches an object.