REST API v1 and v2 of the datasets and their files, out of dataset: fine token scopes, public reads of what is published, tokens limited to some datasets, versioned schemas and the audit trail of every write; v2 with public ids, cursor pages, problem+json, ETags and publication as a resource
The page of a hub's search: /explore, /api/v1/search with its declared schema and the header search box, over the public sources the features register in the framework's search registry (a hub's datasets and files, their facets and orders), which the search feature browses with Elasticsearch or in memory. It names no kind of material, so another hub explores its own.
Zenodo as a publication provider of dataset: deposits a hub's datasets and gets their DOIs (the sandbox outside production)