Hi,
I have not understood the meaning of advertised BBOXes from the remote server. For what Geoserver is using that information?
I believe that most WMS servers do not advertise bounding boxes for all the supported projections, the most famous example is just Geoserver http://demo.opengeo.org/geoserver/wms?service=WMS&request=GetCapabilities&version=1.3.0. Mapserver by default is advertising also BBOX only for the first supported SRS but it can be extended by a configuration option.
I have not thought thoroughly but my feeling is that WMS clients should not care about advertised BBOXes but that is probably because I run Mapserver which believes whatever layer extents I manually write into the mapfile and they are systematically inaccurate and often totally wrong, I fear. WMS servers send empty maps very fast if WMS client is sending BBOX which is outside the data area. However, I can imagine that sometimes, for some servers, there can be more trouble if client is sending a BBOX-SRS combination that does not work because re-projection with those values is impossible.
I suppose that the native SRS has meaning only when Geoserver must re-project map images on the server side when client asks for such SRS which is not supported by the remote server. Otherwise it is forwarding GetMaps by using the same SRS that was used in the client request. And when server side re-projecting is needed it would be best for the image quality to fetch maps from the remote server in the native projection of the data for preventing double-re-projecting. Often the native SRS is the first advertised SRS in GetCapabilities and therefore I think it would be better to use the first advertised SRS of the remote server as the native SRS than EPSG:4326. At least the quality of all Finnish and Norwegian raster maps would suffer badly if they would be squeezed into EPSG:4326 before the second re-projection.
-Jukka Rahkonen-
Tore Halset wrote:
On 29 Sep 2014, at 09:10, Andrea Aime <andrea.aime@anonymised.com> wrote:
> We could back transform each bbox to wgs84 and pick the largest one,
> yet, the reprojection itself back and forth can make bboxes larger
> each way one goes around (because of the rotation and the fact that
> the tilted reprojected bbox is turned into another axis parallel
> bbox), so reprojected bboxes migth look larger than a "original" wgs84 one.
>
> I'm leaning towards keeping things simple and just pick the first bbox
> in the list. Would that work in your case?Picking the first is fine for me as I can tune the capabilities myself. However,
choosing the box for EPSG:4326 would be a more logical (IMO) choice. And
layer.getBoundingBoxes() return a HashMap, so we will have to dig a bit deeper
to find the *first* than just adding a "break;".Regards,
Tore Halset.------------------------------------------------------------------------------
Slashdot TV. Videos for Nerds. Stuff that Matters.
http://pubads.g.doubleclick.net/gampad/clk?id=160591471&iu=/4140/ostg.clktr
k
_______________________________________________
Geoserver-users mailing list
Geoserver-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/geoserver-users