Talk:GSoC Applications 2008

Routing

There are several proposals for routing, some more generic and some more focused on particular modes of transport. Can these be combined somehow? Ideally we would have a fairly extensible routing solution that could, for example, do public transport routes, official cycle path routes, walking routes that go through green space, etc. TomChance 10:10, 8 April 2008 (BST)

I would be interested in plugging a routing solution into a slippy map I'm working on. I would prioritise routing applicants who could make it quite generic, so anyone with an OpenLayers site and access to the database (if needed) can roll their own routing interface. TomChance 10:10, 8 April 2008 (BST)

Special care should be taken that we don't copy data from Geonames wholesale because that data could be "polluted" with Google copyright. Just adding foreign keys (like "geonames:id") to already existing objects should be safe tough. Hakan 11:02, 8 April 2008 (BST)

OSM data should be always be regarded as more specific. When in doubt, accept the coordinates in OSM instead of synchronizing with geonames Hakan 11:04, 8 April 2008 (BST)

We should only sync data that can be proved to be not from Google Maps. That could be elevation data, population (which GN itself gets from another service calls "World Gazeteer"), alternate names, ...

How many times should a tile be rendered: one per language defined in the system, or one per language defined in the zone? How to define zones (countries): polygons, or tile sets (like osmarender's z12 land/sea)? Ivansanchez 11:47, 8 April 2008 (BST)

Let me put it in this way that, a tile/zone will be rendered separately for a particular language defined in the system, only if its' tranlations are available for that language at that point of time. Otherwise the default renders could be used in areas where localization is still not done. And zones can be defined as tile sets, typically at different zoom levels of osmarender. It would be great to have suggestions about more optimized approach. ArindamGhosh 02:38, 10 April 2008 (IST)

I would suggest a phased approach. First add system configuration to associate a users language with a tile set. A pre-existing tile set to test this against would be http://freemap.sk/ and the Slovak language. Next look into generating tile sets based for each local language. There's a couple possibilities here -- have mapnik generate entirely new tiles for each language, or seperate the geometries from the labels and generate semi-transparent label tiles seperately. The second possibility is probably going to be more efficient in the long run. As for the zones -- I'm unclear if this is necessary. Isn't this effectively what the default 'name' key indicates? In any case, I think we'll need to evaluate during the project how much of this task is possible. Just doing i18n/l10n within rails is going to be a good chunk of work. Mikel 15:30, 11 Apr 2008 (PST)

This seems much better. Testing on existing tile sets is a good idea. And its' better to follow the more efficient method to generate tile sets for different local languages. Yeah, I think the we may not create separate zones as such because default 'name' key does the job quite well. ArindamGhosh 20:50, 12 April 2008 (IST)