Anything can be done in software, but you would make it difficult for yourself.
In summary, the JSON-RPC API focus on WebUI interaction (used by a human being) with ConfD, while the RESTCONF API target REST machine-to-machine point-to-point communication.
Just browse the man page for the JSON-RPC API for 5 minutes and you will find things like for example:
- JSON-RPC support the candidate datastore. RESTCONF does not. (e.g. pre-validate configuration in the candidate before a commit to the running datastore)
- The cookie based authentication that creates sessions and transactions in sessions. RESTCONF has no session concept, so each PUT/PATCH/DELETE/YANG patch will be a commit, while with JSON-RPC you can make changes and not until you commit the changes in the transaction are they applied to CDB. And if the session or transaction fail before commit, nothing is set. (for example a commit button that you can press after you are done configuring something)
- You get the permissions for users with JSON-RPC (for example grey fields and buttons when a user does not have privileges to change or execute something)
- WIth JSON-RPC you get access to the complete schema as generated by the YANG model and you can get individual levels, which I demonstrated earlier. (download only the config users view on their current screen)
- rollbacks, messages to other WebUI / CLI users, subscribe to changes in config and oper datastores, get transaction changes (diff) etc.