[OSM-dev] how many api calls/time for an app allowed?
Roland Olbricht
roland.olbricht at gmx.de
Fri Jul 3 16:12:08 BST 2009
Hello,
> i am playing with the idea of providing "real time" error checking on
> osm data. usually one would use the XAPI. but since it is unreliable and
> obviously misses data and it seems no one really cares (no ansers to
> questions about that matters on the mailing list - at least not
> satisfying ones) one would use the API instead.
Well, one of the design goals of OSM3S was to provide an almost real time
error checking. Pleas have a look at
http://78.46.81.38/#section.debug_area
to see how to find bugs in the borders of areas with OSM3S. This is just a
showcase.
The big idea to project is after is to divide between the code to query the
database (currently C++) and the code that describes the error checks
themselves (a scripting language specifially crafted to support advanced OSM
data queries and error checking rules). Thus the former can be ugly but
performant. The latter should be human readable and so safe that it can be
directly submitted from users through the internet (as opposed to arbitrary
SQL queries). Due to the complexity of the computations (anyway some hours
for the entire planet), I'm aiming to be perpetually 4 to 6 hours behind the
main API. The concept can be developed to be only 60 to 90 minutes behind
the main API, but this needs a lot more effort. The project is still in an
early stage, e.g. the documentation of the OSM3S API is not yet complete.
I would be glad to share efforts. This includes developing the scripting
language towards the features it really needs, as well as improvements to the
database source code (by the way: what is an appropriate way to publish the
source?). I don't insist on C++ or the scripting language syntax or
semantics; I have chosen it because the tools for it are at hands. I would
also like to discuss the entire concept as it is a test case for a larger
project.
Cheers,
Roland
More information about the dev
mailing list