[OSM-dev] OSM API Improvements work

[Paul Hartmann]

On 10/11/2012 09:26 PM, Tom MacWright wrote:
> Hey dev,
>
> Just posted the first few issues of the work that I can deem 'stuff that
> we're doing as part of the Knight iniative'. They consist of API-related
> tasks, some of which have had prior art but haven't been tested/completed
> enough to ship. I'd like to get them done and shipped to make some
> substantive improvement in the API.
>
> The first three are:
>
> * JSON formatting for API calls and GeoJSON for some of them. Basically
> just making things friendly for Javascript and other JSON-era languages.

For a Javascript editor, CORS headers are very crucial, it would be 
great to have them implemented for the main API! (Otherwise you have to 
route all the communication to the API through your own server. I.e. the 
download/upload URL must be from the same domain as the Editor URL[1]).

JSON is nice to have, but not an absolute requirement - after all, you 
can parse and serialize XML in Javascript, if you try hard enough. The 
JSON code for cgimap is basically implemented (in case you don't know 
already):

<https://github.com/zerebubuth/openstreetmap-cgimap/commit/914f09c4adfe73535da09b4ff04468a2824d48cf>
(and later commits)

As I understand, there needs to be at least some testing and an 
equivalent implementation in the Rails Port, before it can be shipped. 
(Not 100 % sure about that, you should get in touch with the admins.)

> * Filtering the API endpoints, so that POI editors don't have to sift
> through road data, and so on.

In the open source world, people often prefer to start their own new 
project rather than to contribute to an existing one. Consequently there 
are many small one-man projects that are discontinued after a couple of 
years.

In OSM, this is no different: there are often multiple tools for pretty 
much the same purpose. This is Ok to some degree, but obviously you 
would progress faster, if people worked together on a single project for 
the same amount of time.

That said, Overpass API does a pretty good job at filtering data. So if 
your mission is to improve OpenStreetMap as a whole, in my opinion this 
means improving existing tools, and only fill the gaps if there really 
are gaps.

(Btw.: Overpass API got funded by the FOSSGIS e.V., they pay the server 
costs for 2013 and 2014, so the service should be pretty stable.)

Best, Paul

[1] http://en.wikipedia.org/wiki/Same_origin_policy