No requirements for GET /service_types

[soxofaan]

The service entries in GET /service_types currently have no required fields: each of configuration, process_parameters and links is allowed to be missing.

This caused the validator to accept the 0.4-style version of this endpoint in the 1.0-version of the VITO backend, e.g. currently at http://openeo.vgt.vito.be/openeo/1.0/service_types:

{
    "WMTS": {
        "attributes": [],
        "parameters": {
            "colormap": {
                "default": "YlGn",
                "description": "The colormap to use.",
                "type": "a colormap to apply to single band layers"
            },
            "version": {
                "default": "1.0.0",
                "description": "The WMTS version to use.",
                "enum": [  "1.0.0"  ],
                "type": "string"
            }
        }
    }
}

Note these errors (that the validator didn't catch):

• 0.4-style fields attributes and parameters
• invalid "type": "a colormap to apply to single band layers"

Can this be avoided in some way?
For example require that process_parameters is present with at least and empty array?

[soxofaan]

cc @bgoesswe

[m-mohr]

Due to the "extensibility" of JSON, it's hard to avoid.

I'd actually like to require configuration and process_parameters as it hopefully makes the back-end think what it needs to set ;-) and it's more explicit. Not sure whether we should introduce the breaking change in the release week though with just the reason to help the validator.

It doesn't catch the invalid type as it's in an "undefined" structure, but it would be caught if it's in process_parameters.

[m-mohr]

What I guess would be useful to have in the validator, that it reports "undefined" properties as warnings or so. @bgoesswe

[soxofaan]

Not sure whether we should introduce the breaking change in the release week though with just the reason to help the validator.

according to the openeo hub the support for secondary services is pretty thin (and so I assume practically unused), so I don't think a lot would break actually. But indeed it's already late to do these kind of things

[m-mohr]

Indeed, just VITO and GEE. I'm fine with it and if VITO is fine with it, too. It really is not very useful to have nothing required at all.

[bgoesswe]

What I guess would be useful to have in the validator, that it reports "undefined" properties as warnings or so. @bgoesswe

Yep that seems like a good idea, I added a new issue at the validator for that.

[soxofaan]

if VITO is fine with it

yes, it's not being used in a real application or use case at the moment, so we're open for minor breaking changes

[m-mohr]

See PR #322.