It would be desirable to support a fallback service, which serves LOD at a known configurable URI, and contains “trustable” assertions about the licensing of datasets. In the event of a dataset not containing licensing metadata itself, this resource can be consulted.
It should be treated as a crawled resource once it's been fetched the first time (and so hashed in the same way)
As such, it will be refreshed by the crawler's normal processes on a periodic basis.
Similarly, once it has been fetched, any consultation with the resource will be by way of loading the local copy in the cache.
The first-time-fetch process will need to be implemented with care to ensure it is robust in the face of parallelism.
The synchronous 'fetch fallback licensing resource' implementation will need to understand how to follow redirects (whether they're cached redirects stored in the queue database, or new redirects which should be added).
The eventually-used licensing data will need to be tracked alongside the resource being processed so that processing engines (e.g., Spindle) can relay the relevant statements and their origin.
VoID may be a suitable vocabulary for this resource, thanks to void:uriRegexPattern, but consultation may be required; the specifics of this will be dependent upon who, if anybody, is willing to operate such a service.
It would be desirable to support a fallback service, which serves LOD at a known configurable URI, and contains “trustable” assertions about the licensing of datasets. In the event of a dataset not containing licensing metadata itself, this resource can be consulted.
void:uriRegexPattern, but consultation may be required; the specifics of this will be dependent upon who, if anybody, is willing to operate such a service.