[OSM-talk] Responding to vandalism

Tomas Straupis tomasstraupis at gmail.com
Fri Mar 17 15:50:53 UTC 2017


Let's get on the higher level first.
There are two ways of doing it from the process perspective:

1) EDIT->TEST->COMMIT
2) EDIT->COMMIT->TEST

The first one gives higher quality but also discourages edits and
maybe even prohibits edits in areas with no/few "checkers".

So obviously the way to go for OSM is option 2.

Now here what has to be done is an appropriate testing mechanism.
There are some functionality already done (like the one in Belgium),
but the problem is that everybody sees ALL last changes, there is no
way to SHARE the work of checking and you never know if somebody has
already checked the changeset.

What we are doing in Lithuania for the last 5 years or so is we have a
patrolling mechanism similar to wikipedia. That is all changesets in
the region (in our case in Lithuania) are filtered out and placed into
"check list". If the editor is known good mapper - his/her edits are
"approved" automatically. Otherwise somebody with a status of "known"
mapper should approve it. But when the changeset is approved - it does
not show up for other "approvers". This way we avoid double work. So
in practice this allows us to review only "suspicious" changes and in
5 year of experience this worked out perfectly - all bad/suspicious
changes have been noticed in a matter of hours! (for example all
suspicious crap.me edits can be reverted promptly)

So my suggestion is to add some global "patrolling" mechanism with
division to regions (maybe by countries, maybe by country regions for
large countries). So if there are people interested in some region,
they will review the changesets, if there is nobody interested -
nobody will review, but changes will be in database anyway - so no
preventing of edits.

P.S. Second step would be more automated checks but that is a separate
topic and should only go after this first one is solved/implemented.

-- 
Tomas



More information about the talk mailing list