[OSM-talk] Need help!
frederik at remote.org
Fri Sep 12 13:22:22 BST 2008
this is my report on undoing the changes made by user "Dmitry
Olyenyov" on 9/10 September.
A total of 76.109 nodes and 8.919 ways were edited in that time span by
Of the 76.109 nodes:
- 73.724 were modified or deleted with Dmitry's being the last edit
- all of these were restored to the previous verision by the undo
process. The people who had last edited these nodes before Dmitry
(and whose version has been restored) are: 24.718 AkMeR, 17.481
alexfrol, 14.071 katpatuka, 9.070 Max Vasilev, 1.514 Digy, and
44 other users with less than 1.000 edits each.
- 1.642 were created
- of these, 713 had been deleted already by someone else when my
script went to work.
- 250 could not be deleted by my script because they had been
incorporated into ways that were NOT reverted, and so could not
- 679 nodes were deleted by the script.
- 743 were modified or deleted with someone else having edited them
after Dmitry (734 AkMeR, 9 Johnny Carlsen). These were not touched
by the script.
A further 2.063 nodes had been deleted by *other* people (mostly
"alexfrol" it seems) in the mean time, but were needed to reconstruct
the ways to the state they were in befor Dmitry touched them. So I
undeleted these nodes. (It seems that Dmitry's actions may have left
these nodes sitting around unconnected somehow, and then someone else
came along and removed them?)
Of the 8.919 ways,
- 487 were created by Dmitry; 479 of these were deleted by the
script. 3 had been deleted by someone else already, and 5 could
not be deleted as they seem to be part of relations:
- 7.947 had been modified or deleted by Dmitry and were reverted
to the previous state by the script (previous editors: 2.406
AkMeR, 2.002 katpatuka, 1.510 alexfrol, 540 Digy, plus 49 more with
less than 500 edits).
- 81 had been modified by someone else in the mean time and were
not touched by the script.
- 404 could not be reverted right away because some nodes for them
were missing; after re-instating the 2.063 nodes mentioned
above, all of these ways could be reverted to the pre-Dmitry
version as well.
I hope that while a few isolated spots of trouble remain, this has
remedied the bulk of the problem.
I would like to find out more about how exactly this happened, to
perhaps improve our tools to avoid such accidents in the future.
I can provide detailed lists of the object IDs for any of the
aforementioned groups in case someone wants to do manual checks.
@AkMeR: I hope your work looks right again now - you are a very prolific
@Dmitry: Don't worry, shit happens. Tell us what exactly happened.
More information about the talk