Modeling a read-only singleton is more straight-forward than modeling a parameterless function and a function import.

Also the update restriction may be compatibly relaxed in a later schema version, whereas switching from a function+import to a singleton is an incompatible change.