[OSM-talk] Revert a changeset

[Peter Miller]

On 18 Aug 2009, at 20:08, Teemu Koskinen wrote:

> On Tue, 18 Aug 2009 20:48:03 +0300, andrzej zaborowski <balrogg at gmail.com 
> > wrote:
>
>> 2009/8/18 Peter Miller <peter.miller at itoworld.com>:
>>> On 18 Aug 2009, at 14:57, Teemu Koskinen wrote:
>>>> Could somebody please revert this changeset:
>>>> http://www.openstreetmap.org/browse/changeset/2168210
>>>>
>>>> The moving of the nodes across the Atlantic is obviously wrong.
>>>
>>> Do check out this page for guidance and the email address for  
>>> requests
>>> to the data working group.
>>> http://wiki.openstreetmap.org/wiki/Vandalism
>>>
>
> I don't think this case was deliberate vandalism, other edits from  
> the user seems to be good.

An important distinction. The vandalism page does already make the  
distinction but no one is disputing the need to revert such a change- 
set.
>
>>> Note that I have been working on this page today and have added a
>>> section for 'speedy response' in cases where a failure to respond
>>> within hours could lead to highly visible damage to the rendered  
>>> maps
>>> or changes in sensitive areas (for example Washington - particularly
>>> sensitive given the support and visibility given to OSM by the
>>> Whitehouse).
>>
>> Note that most incorrect edits spanning more than a few nodes need a
>> speedy response because soon people start making edits on top of the
>> unwanted changeset and reverting it becomes more difficult.
>>
>
> What we need, as has been previously discussed on the list, is a  
> similar mechanism that wikipedia has that will revert an edit  
> easily, maybe even from the website ui.

Agreed. It is worth noting that a lot of the anti-vandalism tools and  
general data improvement tools for Wikipedia have been developed and  
are made available independently from the main project. Undo is core  
to the Wikipedia project and I think we need such a button to revert a  
change-set in the UI. I would support the inclusion of undo and  
rollback into the general toolset, but possibly have the feature only  
available to 'established' users, ie ones who have made >x edits over  
 >y days. Wikipedia has some functions, such as image upload and edits  
to much-vandalised pages, that are limited to 'established' users.  
Possibly it is the same concept and definition of being 'established'   
that makes one eligible to vote at the AGM.

>
>> Since I had the setup for this ready, I reverted the changeset  
>> 2168210
>> in my changeset
>> http://www.openstreetmap.org/browse/changeset/2192016
>> but I had to make a couple of edits before uploading it:
>>
>> * xybot had helpfully made an edit on top of some of the nodes
>> removing a spurious tag and causing conflicts.
>> * I did not revert the creation of node 469327157 (a parking) which
>> seems genuine.
>> * Something really strange: node
>> http://www.openstreetmap.org/browse/node/270798013/history is edited
>> two times inside the same changesets and revert.pl didn't deal
>> correctly with this.
>>
>
> There still seems to be some problem, the way http://www.openstreetmap.org/browse/way/39175980 
>  still goes across the Atlantic, but it looks different than before.
>
>
>>>
>>> Personally I think we need a huge effort to be ready for damaging
>>> vandalism and much better tools to spot potential errors in a much
>>> more sophisticated way.
>>
>> Agreed.
>>
>
> I spotted this with the Geofabriks OSM Inspector, but that's still a  
> bit too slow to update, it would be much better if it updated at  
> least hourly or even from the minute diffs.
>
> The revert tools should also be made to look what exactly was  
> modified in the changeset. Eg. if a node was moved, but tags were  
> left untouched, and after that someone else modified only the tags  
> but didn't move the node, reverting the first change should only  
> move the node back to it's original position and not change the tags  
> back as those were changed by someone else.

Agreed - I see no reason why people can't write tools that use the  
minutely diffs and monitor for edits by people not on a 'white list',  
that are possible block shifts of many nodes, that have text fields  
that contain dubious content, that have very long ways, that include  
un-tagged ways (santa-trails), name changes for well established  
features (ie roads that have been called 'High Street' for 2 years and  
suddenly become something else, broken areas etc etc.


Regards,



Peter


>
>
> Teemu Koskinen