Selon Jukka Rahkonen <[email protected]>: > Even Rouault <even.rouault <at> mines-paris.org> writes: > > > > > So it is WFS 1.1.0 request with the BBOX filter. But is it OK to give > BBOX > > > as GML2 gml:Box if the service is WFS version 1.1.0? I have been thinking > > > it should be using GML3 and gml:Envelope instead. > > > > I suppose you're right. I didn't recall all the details, so I had a look at > > the code, and here's the answer : > > > > /* This is a heuristic to detect Deegree 3 servers, such as */ > > /* http://deegree3-demo.deegree.org:80/deegree-utah-demo/services */ > > /* that are very GML3 strict, and don't like <gml:Box> in a > <Filter><BBOX> > > */ > > /* request, but requires instead <gml:Envelope>, but some servers (such > as > > MapServer) */ > > /* don't like <gml:Envelope> so we are obliged to detect the kind of > > server */ > > > > So basically it seems that most servers are tolerant and will accept > gml:Box. > > But some servers that should accept gml:Envelope don't. So the default > > behaviour is to use gml:Box, except in cases where we know it won't be > > accepted (but the server mentionned in the comment is no longer online, and > I > > cannot find another one that doesn't accept <gml:Box>) > > > > It would perhaps be more reasonable to implement the reversed logic from a > > theoretical point of view, but unless you can point me to a server that > > doesn't like the current logic, I'm quite hesitant to change that > > Hi, > > The deegree3 demos can be found here http://demo.deegree.org/ and the Utah > demo > seems to be at http://deegree3-demo.deegree.org:80/utah-workspace/services > I had a quick test and it did not complain about using <gml:Box> in <Filter>. > The result looks like correct, too. So perhaps we can conclude that using > <gml:Box> with WFS 1.1.0 filters is probably wrong but because it works then > it > will not be fixed. > > I do not really like this ambiguous behaviour, though. Now for example WFS > 1.1.0 > bounding box guery with GDAL and EPSG:4236 leads to a mixture of using GML2 > <gml:Box> that does not exist in GML3 but with a GML3 style Northing-Easting > coordinate order. However, normal GDAL user does not ever see it without > sniffing the web traffic and there is really nothing new with WFS being > ambiguos > so perhaps I just need to take everything that is added to the stew and try > to > remember all the tricks and pecularities while waiting for some comfortable > vector delivery service to appear.
Jukka, I'm just curious : is there an reason for you to look at what exact requests the OGR WFS driver sends to the server ? Does it cause practical problems with a server I would not have tested against ? If so, that might be worth reconsidering the implementation and make gml:Envelope the default for WFS 1.1 (and in case of a buggy server fallback to gml:Box if needed. I should have noted which server didn't like gml:Envelope ...), instead of the current logic which is the reverse. Best regards, Even _______________________________________________ gdal-dev mailing list [email protected] http://lists.osgeo.org/mailman/listinfo/gdal-dev
