dct:coverage on dcat:Distribution does not conform to DCAT-AP

[sabinem]

Property: dct:coverage
Class:dcat:Dataset and dcat:Distribtution
Conformance Problem:

• on dcat:Distribution there is no DCAT property to replace dct:coverage. But this is because Distribution properties all relate to resolution, format, size, language, etc. that is to properties that refer to the same content in different physical representations. On opendata.swiss dct:coverage is used to model dataseries as distributions of a dataset. But this considered bad practice in DCAT-AP, see dcat:Distributions Usage note should not encourage the usage of Distributions as Dataseries #112

Proposal:

• Gradually phase out dct:coverage on dcat:Distribution and encourage the use of dcat:DataSeries (https://www.w3.org/TR/vocab-dcat-3/#Class:Dataset_Series) as soon as it is ready to be added to DCAT-AP CH. This has to be done in correspondence with dcat:Distributions Usage note should not encourage the usage of Distributions as Dataseries #112

Changed Proposal: Watch what happens on DCAT-AP and DCAT regarding the modeling on Series of Data, since this seems to be a topic where there is some movement in the DCAT community.

[Juan-Juan-1]

Isn't it a opendata.swiss issue too...? What would be the consequence of removing this property? I am convinced it should be removed, but we should also think about the consequences..

[metaodi]

The reason that dct:coverage exists on the dataset is the lack of support for polygons in [dct:spatial(https://dcat-ap.ch/releases/2.0/dcat-ap-ch.html#dataset-coverage):

It is not possible with dct:Location to provide a polygon as geographical coverage, therefore dct:coverage should be used for this purpose (§ 7.3.14 Property: coverage)

So this part:

In this case the DCAT properties dct:spatial and dct:temporal are able to completely replace dct:coverage on dct:Dataset.

is not true.

Of course we could decide that the range of dct:Location (that dct:spatial provides) is enough and arbitrary polygons are not necessary. But I think it's important to have the full picture why this property was added in the first place.

[metaodi]

In #61 it is stated that dct:spatial now supports polygons (via GeoDCAT). If this is the case, forget my previous comment and get rid of dct:coverage 😁

[andreasamsler]

I would need an example of a Dataset/Ressource at hand to be able to understand from a user's pespective what need(s) we want to cover with this change.

[Juan-Juan-1]

Hi @andreasamsler, in this case there wouldn't be a big change for the user. The catalogue would be structured slighlty differently, so that the information may (it depends how opendata.swiss would implement this change) be presented in slightly in a different way. It's way more a compatibility issue with DCAT-AP: this value does not exist there and this kind of information is carried through dct:spatial and dct:temporal instead. This may lead to issues like the information is not made available on data.europa.eu.

(Note that I'm not taking into account the topic "dcat:Distributions Usage note should not encourage the usage of Distributions as Dataseries #112", since it's only indirectly linked to it).

[sabinem]

@metaodi and @andreasamsler I have launched a discussion about this topic of modeling a Series as Distributions on DCAT-W3C: the suggestions there are quite good. Feel free to jump right into that discussion: w3c/dxwg#1429

For dct:coverage on dcat:Dataset I created a new issue, since this case is more clear and can be decided already, whereas
the case of dct:coverage on dcat:Distribution is harder to solve until DCAT and DCAT-AP over support for the related use cases.

• New issue for dct:coverage on dcat:Dataset: remove dct:coverage form dcat:Dataset as it can be replaced by dct:spatial and dct:temporal #189
• this issue's title and descriptions were adjusted, so that they only relate to dct:coverage on dcat:Distribution. I also added a changed proposal. Please let me know what you think about this: should it replace the original proposal?

[metaodi]

I think we should include this proposal in 3.0.0 to
a) make clear dct:coverage should not be used for DatasetSeries
b) make sure #214 is adopted as well, which actually add the new DatasetSeries class

[metaodi]

Update: I think we should close this issue and use the proposal in #112 instead.