[Imports] Importing Arkansas data

[Michael Leibowitz]

On 04/04/2011 10:56 AM, Ian Dees wrote:
> On Mon, Apr 4, 2011 at 12:41 PM, Michael Leibowitz 
> <michael.leibowitz at intel.com <mailto:michael.leibowitz at intel.com>> wrote:
>
>     If one were to make a one-time import, not only would they
>     probably stomp on a lot of entries already existing, the data
>     would become wrong eventually.  However, a continual import has
>     value.  Although survey is good, the GIS department's opinion of
>     address is the canonical source.  Likewise for other data sources.
>
>
> Ah, this last sentence here is where our problems come from.

It's intended to be a blanket statement.  I don't actually mean that 
import source trump manual survey.  However, there are types of data 
that are not readily observable by survey and have a canonical source.
>
> What is the purpose of a (Street)map that is Open if all we do is 
> import the "correct" data? The Europeans (who are oddly silent in this 
> debate *nudge Fredrick* :) ) will tell you that they don't want the 
> project to be reduced to a bunch of human conflict resolvers. They've 
> generated oodles of data that in many cases is *much* better than the 
> "correct" data provided by municipalities -- all without major imports.
I would posit that in the US, imports have been important.  However, 
they have had problems as you pointed out previously.  Couldn't it be 
the case that imports can be better prosecuted with better tools such 
that their utility goes up and their source of problems goes down?  I 
think that's really what I'd like to see.
>
> Because of this, they will tell you that continual import has no 
> value. In fact, they'll say that continual import has *negative* value 
> because it actually pushes mappers away. Why would anyone sign up to a 
> project where, after they nudge the previously-imported address point 
> for their house over a few meters, it is accidentally blasted away by 
> an import?

It should never be that user generated content is blasted away for no 
reason.  It is reasonable for expect that in such a case, there would be 
careful resolution of such a conflict.  In some cases, the import source 
is wrong (Tiger, for example).  In some cases, the user generated survey 
data is wrong (it does happen).

>
> Yes, in the US there are *tons* of existing datasets that are 
> maintained by municipalities, but we still need to figure out the best 
> way to use that data. For now we need to take it on a case by case 
> basis because of imports that don't quite get the process right (like 
> the one that started this thread) and because of imports that don't 
> add valuable data to the OSM dataset.

While I think parcel data has merit, I'm in no way defending the way 
this import was conducted and think you raise all valid and true points.