IMO any project using #RDF and #SemanticWeb should have a path to exposing their data as Triple Pattern Fragments so that it can be actually usable. Without it we get into a situation where everyone needs to duplicate everyone's data and increase resource usage. It also cements large centralized databases into power. Having a lightweight query layer is essential for having lightweight clients.

https://linkeddatafragments.org/specification/triple-pattern-fragments/