[OSM-dev] Server-side data validation

Paweł Paprota ppawel at fastmail.fm
Fri Jul 13 22:55:07 BST 2012

Hi Peter,

Thanks for taking the time to respond. There was a discussion on
#osm-dev around the whole validation thing, good points were made on
behalf of the "free tagging and no validation" approach.

My proposal would be some middle ground between current state where
"anything goes" and hard validation where Mary Mapper as you mentioned
cannot enter something new. I believe there is a way to provide kind of
a lightweight structure or process around inventing new tags/values that
won't kill the freedom but still allow for consistency/sanity checks.
What that would be, I cannot say right now to be honest - maybe some
form of voting or survival of the fittest tags/values.

However, during the IRC discussion I realized that for me more important
than the purely syntactic issue of tagging is the _semantic_ validation
and quality of the data. Now most of the QA tools work on a very low
level - case of duplicated nodes or stuff like that. I think it would be
good to provide more insight into OSM data, something like comparing the
OSM data with reference sources could go a long way to improve
collaboration experience and data quality in effect.

For example: take a list of places (cities, villages) as a reference
source and use it to verify if OSM contains those places, visualize this
information, enforce it - if someone deletes some small village (if
someone removes New York then it will be noticed of course) - present
this violation in a way that will lead to fixing the situation quickly.

Do such things per country/region + present to the user locally relevant
reports such as missing places from example above and I think it could
only bring benefits and more engaged collaborators.

All this ties in nicely with the usability discussion - not only there
would be technical challenges with implementing such "semantic
validations/reports" but presentation/interaction challenges with making
this actually usable for the user.

I don't know if this makes sense to anyone. Right now I don't have
specific vision how this might be implemented but maybe there's
something in here is worth the effort... For now I will continue to dig
into the Rails app, maybe pick up some tasks and think some more on
these ideas above.

If you have any suggestions or similar ideas, let me know.


More information about the dev mailing list