[OSM-dev] GSoC Ideas

[Milo van der Linden]

Dear all,

Iván Sánchez Ortega wrote:
> El Miércoles, 25 de Marzo de 2009, Milo van der Linden escribió:
>   
>> Mapnik could act as a wms server, but this would not be good for
>> performance since all rendering would be done straight from the
>> openstreetmap db on the fly.
>>     
>
> You're wrong - mapnik depends on a local PostgreSQL database, not on the OSM 
> API.
>   
You telling me I am wrong may as well be wrong. I think we are 
miscommunicating and maybe we just mean different things. Mapnik is a 
WMS server for a wide variety of GIS datasources, one of them being 
postgis, another shape. Mapnik indeed does not depend on the local API 
because if all mapnik servers around would constantly connect to the api 
to render maps, the api server would be heavily stressed. The current 
Infrastructure where hunderds of mapnik/postgresql/postgis servers pull 
data from the osm api-db, turn it into tiles and serve these to 
thousands of end clients is an efficient structure. I would not like to 
know what would happen to the openstreetmap server if the 100 x 1000 
individuals would directly connect to the central database... So please 
do not try to create a static map generator that depends on the API server.
>   
>> A better trick would be to implement a tile-layer in a wms image server, the 
>> image would then take the required tiles, sticht them together in the 
>> requested projection, cut the area that is requested and return that to the 
>> client. 
>>     
>
> If you're going to do WMS, better to do it from vector data rather than from 
> stitching and warping tiles together. You'll ensure better place naming and 
> consistent widths.
>   
Depends on the speed you need, It may be a good and fast solution to 
stitch existing tiles and serve them through wms giving you more 
dynamics in the image width and height and projections. And if the 
stitching is done in the same manner openlayers as a web client 
stitches, there will be no problems with widths and labels. The main 
reason I am giving my opinion is that I would like to prevent the static 
map project to become a project that will reinvent an existing wheel and 
end up having not enough user base to be really inovative. My guess is 
all students that enter the GSoc want to be doing innovative, new things 
that might "shake up the world"
>
>   
>> Caching with WMS is something that, as far as I know is not common, If
>> you can figure out a way of caching and write a proposal,
>>     
>
> WMS caches tend to not be efficient, so they end up not helping much. You want 
> cached maps? Then use tiles.
>   
**Sigh** Is that not what I am saying? use tiles and stitch them through 
wms. Get the best of both worlds..
>
> Cheers,
>   

Kind regards,

Milo