Publish a model. Test an assumption. Find the data behind both.

A registry for grid data, assumptions and models. It holds what is registered with it, points at what is already hosted well elsewhere, and says which is which.

What the Hub holds, and what it points at

An object registered here — a model, an assumption set, a study, a reference network — is one the Hub promises will not move, so the Hub holds the bytes. A catalogued dataset is a description of something somebody else promises, so the Hub points at it and never proxies it. Custody, not size, decides which: a four-terabyte reanalysis archive is free to describe.