So, every HAF app will be writing custom_jsons on hive, and if the HAF app code is open-sourced, other apps can communicate with that app, given the data is now public.

Yes, that's essentially correct.

In fact, I expect most cooperative HAF apps will directly communicate via a shared HAF database . This means that one app can directly read 2nd layer data generated by another HAF app (when the administrator of the HAF servers sets permissions appropriately). This will enable extremely fast and efficient communication, far beyond what is possible today via RPC calls between apps.