[OSM-talk] Brainstorming: Simple Revert-Tools

[Peter Miller]

On 2 Sep 2009, at 21:52, MP wrote:

>>> be added at the right side of every changeset line. One click might
>>> report "succeed" or "conflict/failed, do it manually".
>>
>> Until this is possible I want to try to (or at least collect all
>> necessary information to) write an external tool for this.
>
> Adding "revert" to main site could attract vandals (ok, let's just
> revert stuff) or experimentators (what does this button do?). I think
> better would be to have it as external tool that can either revert or
> prepare file for reverting (so that it would be loaded, checked and
> uploaded in JOSM) - reverting should be easy, but not that easy, that
> it would suffice to click one button to revert.
>

Only 'established users' can upload images to Wikipedia and I would  
suggest that the revert option is only available to established OSM  
users. It would need to take many edits to get 'established' but it  
means that there are some controls on 'drive-by vandals' who register  
and then cause mischief. I guess one can loose one's 'established  
user' credentials by partaking in vandalism.

>> Maybe such a tool could block such revert-reverts (A revert's Bs
>> changeset, B reverts As revert) and only allow them to others, to  
>> avoid
>> revert-wars. Maybe this should only happen after two rounds?
>
> How do you (automatically) distinguish revert from ordinary edit (or
> "almost-revert" from ordinary edit)? I think you can't at least not
> reliably (you can't rely on the commit message, etc ...)

I am not sure you need to. Not all reverts in Wikipedia use the revert  
mechanism, because for complex multiple edit vandalism it is easier to  
go back to a previoud version and cut the text from there and paste it  
in again. The main thing is to be able to see the differences between  
now and then for an area and decide what sort of revert is  
appropriate. For a simple single changeset a revert might often be  
good, however sometimes one might decide one needed to revert a part  
of the dataset wholesale. Not sure why.

A good tool might be able to identify a revert war by spotting tags  
for features that are flipping backwards and forwards frequently or  
tags that get created and deleted. Basically any feature that returns  
to a recent state again a number of times in a short period.

>
> So for avoiding revert war we need to use same mechanisms as for
> avioding any other unwanted edits - block users/IP's from editing, etc

These are surely the last line of defence and are easily circumvented  
unless the other devices are also in place.

To start with we need to be able to deal with the basic drive-by  
vandal (using a simple revert) and then be able to deal with people  
registering many names and doing a few edits with each (white lists  
and patrols might help here). For revert wars we need detection of  
flip-flopping features and we need human arbitrators. Only if all that  
fails do we need to start banning people which is something which only  
the Data Working Group can do - everything else if for the community  
to organise.


Regards,


Peter


> ...
>
> Martin
>
> _______________________________________________
> talk mailing list
> talk at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk