Open sgrellet opened 5 years ago
EPOS-DCAT-AP allows you to describe services independently from the datasets they deliver (if any), thus also independently from a Distribution
. In other words there is no imposed path Dataset
->Distribution
->WebService
. Presumably the request you mention was done for the sake of the demonstrator. However, you are free to describe only a service if that is your primary focus.
WebService
is a resource and can be linked to a Catalog
with the relationship epos:resource
. At a first glance this seems to be inline with the solution you refer to.
In your case there might be multiple ways to populate EPOS-DCAT-AP with several entrypoints.
WebService
and/or Dataset
- Distribution
and then link to WebService
or even we might consider a mapping with Catalog
(see dct:isPartOf
relationship on Catalog
). In principle all those solutions are feasable, it is a matter of agreements e.g. on the sought level of interoperability.
Let's discuss at the next technical meeting!
As opposed to OGC WMS and WFS, a CSW does not serve a 'dataset'
even a WMS doesn't have to act on a single dataset, but could work on many. A layer could be made up of more than one dataset too.
We have been asked to provide a dcat:Dataset and dcat:Distribution for WP15 OGC CSW service description.
Let's move to lay man terms for the demonstration and not restrict ourselves to DCAT 20140116 and EPOS_DCAT_AP {lay man on} But what is the 'dataset' served by a CSW ? As opposed to OGC WMS and WFS, a CSW does not serve a 'dataset' per se but the sum of the union of all the metadata records and datasets/services description contained in it. {lay man off}
According to dcat 20140116, the previous reasoning would imply that is a need for a direct link between dcat:Catalogue and epos:WebService. And not impose to go through the dcat:Dataset -> dcat:Distribution route
W3C DXWG already discussed and solved the issue here: https://github.com/w3c/dxwg/wiki/Cataloguing-data-services The corresponding pull request : https://github.com/w3c/dxwg/pull/241
Can we reuse the proposed approach ?