[camgunz]

I think the challenge is always the stateless nature of HTTP. It's pretty hard to keep a connection to the API server that has your database cursor.

Everything else has big tradeoffs. You can make sure your rows are sortable by a specific column, but that means your searches must all be sorted by that column.

You can save the results of a search and page through it, but that's a lot of I/O.

You can not do any of the above, but then you run the risk of non-deterministic pagination (I went back and because someone added/deleted a row, that page is slightly different now). You also can't really link to it.

These things might all be fine. In my experience though, you run into people who really want to do search in way A, and will use the tradeoffs for way B/C/D to argue against it, as though A has no tradeoffs.