Do we need a way to distinguish between properties that are omitted due to being aggregated away versus those that don't apply (i.e., in a concatenated result)?

@odata.context allows to identify and distinguish all cases: aggregation result, rollup entities, concat result. The @odata.context describes the properties from the input set that remain in the result after aggregation. Any property of the input/expanded sets not listed here has been aggregated away. In rollup entities, one or more of those properties are omitted. Comparing an instance with the surrounding @odata.context, the absence of a property in an instance classifies this entity as rollup entities, because there is no other way how a property could get lost. Concat combines two aggregated results with potentially different structures as a single result collection. This inhomogeneous structure cannot be described with a single context URL at the level of the collection. If the two partial results deviate structurally, the current JSON spec requires a provider to apply @odata.context at entity instance level for every entity. This is not efficient if done for a large set of such instances, but we do not see any other option with the current capabilities of @odata.context. On the other hand, every part of the result structure is clearly defined, so there's need to further distinguish the cases mentioned in the issue. Close without action Accepted: https://www.oasis-open.org/committees/download.php/51905/odata-meeting-64_on-20140109-minutes.html#odata-577