I agree that RPC is bad, but I don't think it is because it blends the boundary. I think its downfall is that it masks failures.
In this case, code seems to be explicitly labeled as client, server, or both. Sometimes, you really do want both. Consider a calendar which is sent down as HTML from the server for mobile clients, but updated on the fly by javascript in full browsers. Typically, however, your client code is explicitly marked as such and calls into server code with the very proven success/failure ajax callback pattern.
In this case, code seems to be explicitly labeled as client, server, or both. Sometimes, you really do want both. Consider a calendar which is sent down as HTML from the server for mobile clients, but updated on the fly by javascript in full browsers. Typically, however, your client code is explicitly marked as such and calls into server code with the very proven success/failure ajax callback pattern.