[MapServer-WMS]Not returning a service exception when srs in GetMap is invalid for some layers

[nsavard@…]

When the SRS parameter in a GetMap request contains a SRS that is valid for
some, but not all of the layers being requested, then the server shall throw a
Service Exception (code = "InvalidSRS").  (see wms/wmsops/getmap/params/srs/4) 
MapServer WMS needs to satisfy this part of the specification to comply.

Actually MapServer is extracting the value of the "SRS" parameter.  Then it
verifies for each layer appearing in the "layers" parameter of the request, that
the projection string is matching the one asked.  If not the layer is reprojected.

MapServer WMS should at least verify that the projection asked is corresponding
to the one within map object.  If not MapServer should look for a matching
projection within each layer.  If the projection asked is not found within a
layer object, MapServer shall throw a service exception.

[assefa]

Norm,

 I did the changes to this. I made few tests and It seems to work well. It is 
still not commited since It needs additional tests. I have updated xcalibur 
with a new version of mapserv_assefa (in the cgi-bin) so you can do your 
tests. If you want to test it locally on your machine, let me know and I will 
send you mapwms.c.

[nsavard@…]

Assefa,

Yes, please send me "mapwms.c".

[nsavard@…]

Assuming information in bug description is correct and there is a map file that
contains a map object within a specific projection and some layer objects within
different projection than the map object (see mswms_ogc_cite_wms_srs map file).
 MapServer should not throw a service exception when a request is made with a
srs parameter which value is corresponding to the projection in the map object
or in one of the layers.

I copied /path/to/cgi-bin/mswms_ogc_cite to /path/to/cgi-bin/mswms_ogc_cite_srs
in which I modified the "MS_MAPFILE" value to point to
/path/to/ogc_cite/map/ogc_cite_wms_srs.map and I changed the mapserver
executable to "mapserv_assefa".  I then copied "ogc_cite_wms.map" to
"ogc_cite_wms_srs.map" in which I made the following modifications:

advertized projection		"EPSG:4326,4318,4319"
BasicPolygons layer projection to  epsg:4318
MapNeatline layer projection to epsg:4319

I then issued this URL request:
http://server/cgi-bin/mswms_ogc_cite_srs?LaYeRs=OGC_CITE&
VeRsIoN=1.1.1&ReQuEsT=GetMap&FoRmAt=image/gif&WiDtH=168&
BbOx=-0.0042,-0.0024,0.0042,0.0024&HeIgHt=96&StYlEs=&SrS=EPSG:4326

The response to this request is "InvalidSRS".

[assefa]

The wms_srs should read "EPSG:4326 EPSG:4318 EPSG:4319" instead of
"EPSG:4326,4318,4319". I have changed the demo map file and I get a map.

[nsavard@…]

Sorry for that Assefa, you are right.

[assefa]

Fixed according to the bug description. I put the validation code under 
#ifdef OGC_STRICT_MODE. So when building for OGC tests, we need to activate 
that flag. I will update the windows makefile. Could someone do the same for 
the other platforms ?

[dmorissette]

I'll add the -DOGC_STRING_MODEoption to the Unix configure script and Makefiles.

[dmorissette]

Assefa, can we change the compile flag to OWS_STRICT_MODE instead of
OGC_STRICT_MODE? I think that would be more accurate since it would mean "OGC
Web Services (OWS) Strict Mode".

[dmorissette]

Okay, I changed my mind again. I know I'm probably the one who suggested this
strict mode flag, but now that I think about it again, if we have apps that
would break because of that exception then we better fix them since they would
break on other servers anyway. I mean it's a bad idea to let the server operate
in non-strict mode (at least in this case) since that will allow us to build
clients that will appear to work, but will fail with other servers that are
strictly compliant.

So I propose that we remove the Oxx_STRICT_FLAG, and leave this SRS cehck
enabled at all times. What do you think Assefa?

[assefa]

My fear was that many of already setup wms clients would start breaking up when
people upgrade to mapserver 4.4. On the other hand, if we aim mapserver 4.4 to
be ogc compliant, It is the best time to intrduce that and people would
understand the needs for this change. I think we should announce it specifically
to the list. I was actually writing an e-mail to the list on this when I
remebered that we decided to put a compiling flag, so It never went out.  All
this to say that I also think we should remove it.

[dmorissette]

Well,when I was working on this yesterday I realized that (1) that would
probably not affect our client apps since it is MapServer that generates the
GetMap requests, and MapServer already does the right thing (it won't request a
SRS that's not supported for a given layer), and (2) *if* we or someone else
happens to have any client app that doesn't handle SRS properly, then it would
probably not work with WMS servers from other vendors anyway, so making them
break now is just a good thing for interoperability anyway.

I agree that a mail to mapserver-dev couldn't hurt, but I don't think that's as
big an issue as I initially thought.

So are you going to remove the #ifdef or should I do it? Also, a note about this
new exception should be added to HISTORY.TXT

[assefa]

I removed the ifdef and updated the History.txt. E-mail sent to mapserver-dev.

Marking it as fixed.

[nsavard@…]

Verified on Fedora Core 2.  I check the "HISTORY.TXT" too.