Hmm. I think I see why your buffering makes sense now.

>No, I was saying that I have to start from scratch after a crash and that's what I want. I think I was unclear, I only have to check for dupes in the SDF file, not with the master table. That's why I don't want to save any incomplete importation.
>
>>As long as you've got the built-in no-dupes scheme, there's still no problem starting the import from scratch, no matter how it crashed in the middle. Without buffering. But you can do whatever you're comfortable with! :)
>>
>>Wait, are you saying that, if you're buffering, you don't have to start the import from scratch if it crashes? How is that, if it's what you're saying?