[OSM-talk] Zero tolerance on imports

Peter Budny peterb at gatech.edu
Mon Feb 21 16:58:35 GMT 2011

Frederik Ramm <frederik at remote.org> writes:

> Hi,
> On 02/21/2011 04:03 PM, Peter Budny wrote:
>> Those of you who think all automated or semi-automated data
>> contributions are harmful to OSM are dooming this project to never be
>> able to grow to become a leading source of mapping data.
> It is a common fallacy to believe that good map data could somehow,
> magically, be produced from computers that evaluate GPS tracks, camera
> recordings, or aerial imagery.
> If this were possible, then Google et al. would be 10 times as good at
> doing it as we are.

Google, like Waze, has both historic and real-time traffic data
automatically generated by millions people with mobile phones.  So in at
least some ways, they ARE 10 times better than OSM.

> The strength of OSM is the people on the ground. If you try to
> eliminate them from the equation

Whoa, who said anything about eliminating people?  What I'm saying is
that we should find ways to integrate human editors with automated or
semi-automated tools, so that humans can delegate the tedious work to
computers and spend more time doing things that can't be handled by

>> Last year, as part of a school project, I built a robot that will
>> automatically create route relations for all the state highways in the
>> US, being careful not to change or duplicate existing data.
> [...]
>> The code would be in use already if not for a few people running around
>> panicking about my devil-robot and its witchcraft.
> Maybe you haven't been able to demonstrate the added value your
> mechanical edit would bring to the database?

The value is that
http://wiki.openstreetmap.org/wiki/Kentucky#State_routes would show
route relations for all 6000+ state routes in Kentucky, instead of
7... and then I could use the same code to finish the other 49 states in
the US.  And then with minor modifications, I could use the same code in
other countries.

As an analogy, we store OSM's source code in Subversion and Git, and let
those tools compare files when we make a change.  Could this be done by
hand?  Of course.  But why would you want to?  You would produce the
same result (actually, you're more likely to make a mistake than the
computer).  Yes, sometimes the tools come upon situations they can't
handle, and have to let a human intervene, but they relieve us of the
tedious bits.

Some people look at OSM and say, "It needs more tools."  Some people
say, "It needs less tools."  Consider me firmly in the first camp.

> I mean, if it can be
> determined by a robot, then surely it would be redundant to have it in
> the data again?

First, your reasoning is specious.  Consider a shopping receipt: what's
the "added value" to listing a subtotal and total, when these could be
trivially computed by summing the items purchased and subtracting the
amount paid?

Second, the robot's contributions would not be perfect... but then
again, neither are mine.  I've never drive down Kentucky State Highway
483, so any edits I make to it are merely the best I can do given what's
already in OSM.  But if I see tiger:name_base="State Highway 483", I'm
going to put it in a relation with the other ways that match it.  A
robot can do exactly the same thing, only a lot more efficiently than I

And before you counter... no, I don't think it's pointless or wrong to
edit a part of the map I've never been to.  If I (or anyone else) ever
DOES go there, it would be nice to have already improved the map as much
as possible, rather than letting it remain a completely unedited jumble
or void.
Peter Budny  \
Georgia Tech  \
CS MS student  \

More information about the talk mailing list