Hi, On 8 April 2010 12:45, Frank Sautter <[email protected]> wrote: > normally a osm-xml file is piped through a bot (e.g. perl script). if it > stumbles upon something fishy (e.g. a multiple consecutive nodes in a > way) it will correct this problem on this entity. it does not know where > in the world this problem is. it doesn't have a big database in the > background which it could query where this entity is located and store > it until the pipe has reached EOF and then make some handy spatial > chunks of all its changes. > > the next question would be: what is the "right" size of a changeset? > 10m², 100m², 1km², 10km², 100km², 1000km² > > the only solution to really fix the spamming of the history tab the way > the history tab is implemented right now would be to put each change in > its own changeset. but that's what we had in api 0.5 and would lead > changesets ad absurdum and i believe a single bot-run should result in a > single changeset.
It still wouldn't be ideal because you might be editing a long diagonal way and the changeset would be visible in history anywhere in the rectangle defined by the two ends. Really the only solution is if the tab used something like ST_Intersect on the geometries instead of on the bboxes and this is more intensive computationally. Cheers _______________________________________________ talk mailing list [email protected] http://lists.openstreetmap.org/listinfo/talk
