[OSM-talk] Zero tolerance on imports
eda at waniasset.com
Tue Feb 22 12:49:29 GMT 2011
NopMap <ekkehart <at> gmx.de> writes:
>There is a considerable difference between an import into a mostly empty
>area - which is rather easy to achieve and mostly helpful - and an import
>into an already basically mapped area, which is hard to integrate and where
>existing data may be damaged.
I'd hope it would go without saying that an 'import' in an area where the map
is not empty should not be trashing the existing data and replacing it, nor
blindly adding new objects over the top of what's there. It must involve
checking and reconciling the two data sets and making a common-sense choice
about how to resolve conflicts in order to best match what is on the ground.
Indeed, I would avoid the use of the word 'import' at all; it should be
a synthesis of the existing OSM map with the new data.
>There appears to be a common misconception that arial imagery and imported
>data is somehow "better by default" than the hand-drawn traces.
I haven't come across anyone with that misconception, but it is one.
I hope that nobody presumes that hand-drawn GPS traces are 'better by default'
than good-quality aerial photos, however; in urban canyons where GPS accuracy
is low, they are not.
>So, personally, I believe that drawing the first roads on an empty map is
>much more fun and motivation than fixing mistakes in someone else's imported
>data. If the actuality, accuracy and quality of data is less than excellent,
>I would rather not import it at all.
I would say that what matters is not the absolute quality of the imported data
(even an excellent quality map would not be worth importing into Oxford or
Friedrichshain, since the OSM data is also excellent there) but its relative
quality to what OSM currently has. If the existing OSM data is very poor,
perhaps because it in turn came from an earlier poor-quality import (or was
traced from blurry aerial photos), then even a moderately good data set can
add value to the map if used with care. Often third-party data sets lack
detail in some areas compared to OSM, but have much better completeness over
a wide area while OSM has holes.
Ed Avis <eda at waniasset.com>
More information about the talk