GML v3 parser should be more liberal with respect to gml:pos

[tschaub]

The simple features profile specifies that

in all cases, geometry coordinates shall only be specified using the gml:pos for
gml:Point or gml:posList elements for all other types

(See OGC® 06-049r1 8.4.4.10.2.a.)

Our v3 parser is based on the schema referenced in the parser:  http://schemas.opengis.net/gml/3.1.1/profiles/gmlsfProfile/1.0.0/gmlsf.xsd

It can be seen there that LinearRingType cannot contain gml:pos elements.

It could be argued that we should support more than the simple features profile. Before going too far with this, it would be good to decide what exactly we want to support.

[tschaub]

allow gml:pos lists in more geometry types

[tschaub]

I'm against slipping too far from the simple features profile with this parser (I'd rather another subclass that was targeted at a specific version). I'm indifferent about allowing this change.

[ahocevar]

I am against this change. GML was designed to come in different profiles, and usually when referring to GML, the Simple Features profile is meant.

Instead, I would be in favor of creating parser subclasses for other profiles, as proposed by Tim. I think the modular way our parsers are built should make it easy for everyone who wants to support a custom profile to do that. But again, I would be in favor to always create profile against a schema, so people exactly know what the profile supports.

[bartvde]

I agree with not doing this change, same goes for ticket:2330 and ticket:2388 however what naming would you guys suggest to distinguish between a more full-blown GML3 parser and the v3 SF parser?

[christophenoel]

I'm in favor of change. I'm using other viewers and, they don't care about the gml profile. In OGC services, we often have to display GML that is not a simple feature profile.