No problem @luis. I'm confused most of the time.
I did specify unique_by and this does reflect the id I need to work with so that's all good.
I'll give a simplified specific example to better explain my use case.
The data I'm referring to relates to schedules or appointments, the id of which is multipart made up of who and when.
So the record id or 'unique_by' contains but is not limited to the id of the person, and the start and finish times.
So the appointment is created and I send the relevant data to the dataset. What if the appointment changes though, either it gets given to someone else or just changed to a different time-slot. It's a common edit but it also results in a change to the unique_by value. If I just append, I get the new data but the old data remains in the dataset even though it is no longer valid. So I need a means by which to delete the old one either before or after posting the new data.
I understand this would not be as frequent a requirement if the record ids were just ids and only the data was modified but it is what it is. That said, even with sequential ids, there are times when records are legitimately deleted so I think the option is still worth considering.