Nothing in the specification records how much of a content record a retrieval reached. media_type is the medium, not the depth; conformance levels describe which lifecycle stages are reported, not how much of an item was reachable; bounded source portions (#25) record what entered a generation context — a different axis. A retrieval that served only an abstract and a retrieval of full text from which one span was later grounded are indistinguishable at the retrieval layer, and metadata-versus-full-content is the distinction scholarly usage reporting has relied on for two decades.
Proposed, drafted for the v1 release candidate: content_depth on the 6.1 data profile with core values metadata, abstract and full, open values, consumers tolerating unknown ones — the media_type pattern. Three core values because abstract-level access is distinguished in scholarly practice; a sector that needs only two loses nothing. Although listed in the agent profile, the field applies to content_retrieved from any source_role: the origin that served the response knows the depth authoritatively, and origin and edge reporters SHOULD include it where entitlement gates depth. Depth records what was reachable at retrieval, independent of what portion later entered a generation context.
Deferred: whether depth also applies to content_grounded and content_presented, and the mapping to sector vocabularies (COUNTER Investigations and Requests), which belongs in an external profile.
Nothing in the specification records how much of a content record a retrieval reached.
media_typeis the medium, not the depth; conformance levels describe which lifecycle stages are reported, not how much of an item was reachable; bounded source portions (#25) record what entered a generation context — a different axis. A retrieval that served only an abstract and a retrieval of full text from which one span was later grounded are indistinguishable at the retrieval layer, and metadata-versus-full-content is the distinction scholarly usage reporting has relied on for two decades.Proposed, drafted for the v1 release candidate:
content_depthon the 6.1 data profile with core valuesmetadata,abstractandfull, open values, consumers tolerating unknown ones — themedia_typepattern. Three core values because abstract-level access is distinguished in scholarly practice; a sector that needs only two loses nothing. Although listed in the agent profile, the field applies tocontent_retrievedfrom anysource_role: the origin that served the response knows the depth authoritatively, and origin and edge reporters SHOULD include it where entitlement gates depth. Depth records what was reachable at retrieval, independent of what portion later entered a generation context.Deferred: whether depth also applies to
content_groundedandcontent_presented, and the mapping to sector vocabularies (COUNTER Investigations and Requests), which belongs in an external profile.