[OSM-talk] Brainstorming: Simple Revert-Tools

[Peter Miller]

On 2 Sep 2009, at 20:30, Peter Körner wrote:

> Sybren A. Stüvel schrieb:
>> On Wed, Sep 02, 2009 at 09:17:11PM +0200, Peter Körner wrote:
>>> A revert is changing a lot of things in one time, which would be
>>> much more time-consuming with e.g. josm. I see this (doing a lot of
>>> things with just a single click) as the main problem with an easy
>>> revert tool without some kind of vote.
>>
>> In that case the entire API should be scrapped. After all, I can
>> easily make some small program
> You can do as, but with a webservice that takes a changeset-number  
> in a
> form and offers a revert-button, everybody can do this.

Yes, I would like to see a revert function available to which I could  
give a changeset to and it would revert all the changes made to the  
database within that changeset. It must flag the changes as being made  
by me and being made using that tool (I must take responsibility for  
the decision to revert and the tool must also be identified as it has  
some responsibility for the quality of the revert). Would the tool be  
something that ran on my computer? Possibly.

A good tool will need to be able to do this even when some changes  
have been made on top of the changeset and possibly highlight a few  
issues that cannot be reverted because of conflicts.

A good might be able to review all the changes made by a particular  
user over a period of time and list the status of the features before  
and after to assess what the user is doing and if there is any sense  
to it or what. It should be possible to revert multiple changesets by  
one user in this way.

A good tool might be able to monitor the minutely diffs and identify  
unlikely behaviour, such as someone randomly changing names for  
features that have been stable for some time, or moving nodes around  
that have been stable for some time. This is just a warning, not a  
definite problem and would need to be assessed.

A good tool might be able to monitor the minutely diffs and check  
names against a 'swear list' to check for unlikely street names and  
locality names etc. Not all rude names are incorrect as the book 'rude  
Britain' can testify but it is worth checking.[1]  The camp sites at  
Burning Man have some very offensive names as well.

A good tool might be able to import a 'white list' of trusted editors  
and then focus the attention on unknown contributors  in the minutely  
diffs feed. There would need to be a way for trusted users to give  
trusted status to others or challenge it and share lists.

A good tool might spot users breaking coastline or motorways or  
railway lines or administrative boundaries or other very established  
features and highlight this for review.

This tool should be configurable so that one can monitor only a part  
of the world that one is interested in, or only feature types that one  
is interested in, for example railways in Europe, or everything within  
a bounding box.

A good set of tools will allow us to revert vandalism within minutes.

Please can a bunch of coders get on with producing support for this  
important work. We have remarkably little graffiti but it does exist  
and will get more of a problem. It is such a shame to see vandalism  
messing up a lot of good work and we risk loosing established  
contributors unless we can protect the work already done better than  
we are doing.

We will of course have some revert wars, that would be a sign that we  
had the technology and needed to build the social infrastructure to  
control its use. Currently we don't seem to have the technology to  
revert changes where changes have been made on top.

You could check the tool on the edits made over the previous two  
months by Liam123, some of which have still not been reverted for lack  
of a suitable tool to achieve it.

[1] http://www.amazon.co.uk/Rude-UK-Exposed-British-Passages/dp/0752226657


Regards,


Peter



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