I come across logically bad data in OSM on a regular basis.
An example is way 33803055. It consists of a single node.
Another example is way 240561169. It consists of 3 nodes, but with such a little difference, that the length is really 0. There are many such ways, shorter than 0.1 meter.
Why isn’t such automatically deleted or not allowed in the first place?
There are already some services displaying such errors.
In general, see http://wiki.openstreetmap.org/wiki/Quality_assurance
and then http://www.geofabrik.de → Tools → OSM-Inspector → View layer: Geometry → Error type: way nodes
are your errors displayed there?
And maybe other error tools can mark similar errors, maybe http://keepright.at ?
I am looking for automatic clean-up in the master database. Something that runs daily without user-intervention. There are actually lots of stuff that can be deleted automatically. Another example is roundabouts that are also flagged as being oneway. That is not needed and only adds to the size of the OSM files.
Hm. I’m not sure if we really want to do all this completely automatically.
For example, some one-node-ways still contain tags and/or history, which may be reconstructed by experienced mappers. On the other hand, we also have editors that create completely useless (tagless, v1) one-node-ways under some circumstances. But I’d guess that it would make more sense to invest some development time to fix the programms that create the bogus data in the first place. (fix the causes not the symptoms ;))
An additional oneway=yes on a roundabout is indeed superfluous, but it doesn’t hurt anybody and is technically not an error, thus nothing that has to be corrected.