Thanks for the reply. I'll flesh out my thought process in case it's helpful. My immediate reaction was excitement about the abstraction. An example use case is joining my users to their corresponding Stripe Customers in SQL. The kinds of queries I can reasonably write depend on implementation details of the connector. For example, if Stripe has a bulk customer lookup (list of customer IDs -> Customers), and the connector uses it, I can estimate I'd be able to query on the order of 500 Users at a time in a performant way. But if the API only supports looking up one customer at a time, that 500 User query kicks off 500 API requests, which isn't going to work.
You're right -- it's not unexpected -- maybe more like a leaky abstraction.
You're right -- it's not unexpected -- maybe more like a leaky abstraction.