On Thu, Sep 3, 2009 at 3:13 AM, Martin Koppenhoefer<dieterdre...@gmail.com> wrote: > 2009/9/2 Pieren <pier...@gmail.com>: >> 2009/9/2 Peter Körner <osm-li...@mazdermind.de>: >> Revert should >> be possible from the main site everywhere changesets are listed : from >> the history tab on a bbox or the recent changes without bbox ([1]) or >> from an individual contributor ([2]) edits. A new élink "revert" could >> be added at the right side of every changeset line. One click might >> report "succeed" or "conflict/failed, do it manually". > +1 also "already reverted" to prevent the late commers from spending too much time manually finding out that there's nothing left from the bad changeset.
Generalizing a bit: changesets could have a metrics of reversibility: 100%=no changed object was changed afterwards (trivial to revert), 95%=5% of objects were changed since the changeset was done (there will be some conflicts), 0%=all objects were changed (probably a very old changeset or a bad changeset, undone either manually by various users or by admins, or a small changeset where another translation of a city name was added afterwards...). It would be fairly trivial (but still consuming some resources) to keep this metric updated in the database by keeping the number of all objects within a changeset and a counter of changed objects that would be updated every time an object is changed (find changed object's last changeset and increase the counter there). This "Persistance of objects" (POO for short :)) could also be used (as one of the metrics) to determine how established the user is, if needed. The revert link/button should not be placed directly on lists, but should only be accessible after carefully inspecting the new, friendlier changeset diff (see below). >> As a possible watchdog, the system could send automatically a message >> to the author of the reverted changeset like "User:XXX reverted your >> changeset 12345". > +1, that's a good idea > > I would also find it very helpful to have renderings of the different > states on the api-history-pages (of elements way/node), as well as for > changesets (before/after), although this might not be possible in some > cases (e.g. when the area is very big / a way very long). Yes, visualizing changesets (and generally all object history) could be better than it is now: highlighting new/deleted/changed tags (down to individual characters in keys and values and numeric difference for numbers, possibly even calculating the distance and direction of movement for lat&lon), with an aggregated summary of changes. Of course all this wouldn't say that a way or relation changed if one of it's node was moved...should it? Stefan _______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
