Hi, Peter Miller wrote: > there are two ways to removed vandalism [...] the first is to repair > the damage by moving things back to where they should be and > re-entering the correct information in tags, the other is to revert > the data to before the person touched it.
While it would theoretically be possible to do a "proper" revert on a database level, this has to my knowledge never ever been done in OSM and is very unlikely to be done with Liam123's edits. Whenever we "revert" something, we actually retrieve the previous version from the object's history and re-upload that as a new version. Always. You are right in saying that this leaves the person who has done the vandalism on the "history stack". The only way around this (apart from database-level reverts which are extremely unlikely to ever happen) is to delete the object and re-upload the last known good version as a new object. This is however difficult if you have objects referring each other, and also destroys the pre-vandalism history. > Currently people are using a mixture of repairing the damage which > leaving the person in the IPR chain, and reverting ways one by one > which is slow. However... there is too much damage to efficiently do > this manually, hence my question with regard to a programatic removal > of the edits which I understand is possible. I have scripts to do reverts on a larger scale, and even talked about them in a workshop at SOTM so they're not super-secret; but they all operate with normal user privileges, nothing on a database level. What they generally do is just what I described above - retrieve a changeset, look at all the objects, access their history, read the last known good version, and upload it again. Exactly the same as someone would do with the "H" or "U" functions in Potlatch, just automated. I was acutally about to offer my help with reverting Liam123's changeset you mentioned previously but then I saw that most of the changes had been manually fixed by various other users already. > I will ask again. Does anyone support a programatic removal of all > this person's edits? Does any object to a programatic removal of all > this person's edits. Before saying you object please check a > selection of his changes and note that some have been reverted > already so might look ok, but were not ok at the time. > http://www.openstreetmap.org/user/liam123/edits I can, if there is sufficient consensus, revert everything that currently has Liam123's name to it to the state it was in before he touched it (which also includes deleting things he created and undeleting things he deleted). Some things might not be revertable because others have built upon things that Liam123 has added. I could even, which would be more severe, revert anything he ever touched to the state it was in before, no matter whether someone fiddled with it afterwards or not. This is not by special privilege, it is a technical possibility open to anyone with an OSM account and sufficient scripting expertise. However, this is not something that should be done lightly, and the project has not yet evolved ways to deal with this kind of decision-making. How many people would have to express support for such a wipe-out to make it ok? Is it sufficient to ask on the list whether anyone objects or is this something like the "you could have looked at the building plans on Alpha Centauri" in the Hitchhiker's Guide? Who, ultimately, decides what vandalism is, and when all avenues of contacting a user have been exhausted? Should not e.g. the OSMF issue a formal warning shot or something like that? Should the OSMF data working group be involved? All these things need to be considered, especially as we'll have more cases of this in the future and we do not want to set a wrong precedent. Bye Frederik -- Frederik Ramm ## eMail [email protected] ## N49°00'09" E008°23'33" _______________________________________________ Talk-GB mailing list [email protected] http://lists.openstreetmap.org/listinfo/talk-gb
