A discussion that keeps coming up concerns how we should be handling NDRs for our customer installations. It seems that even with the higher-end anti-spam appliances, the “recommended” configuration is to generate bounced messages – even when the incoming mail is scored as decisively spam/uce.
So the question becomes, should we even generate NDRs at all?
Between the fact that incoming spam will invariably have a forged “From:” address, and that some organizations are recommending that we disable all NDRs – why even bother with the NDR?
It certainly doesn’t take too many complaints from your customer’s ISP or the re-addition of their IP/domain name to RBLs to cause your customer have serious misgivings about your service.
That leaves you with disabling NDRs across the board.
The first place to consider disabling NDRs is at the perimeter. So if you have a perimeter antispam appliance, consider not generating NDRs there. If your perimeter happens to be your Exchange server itself, you might want to consider disabling NDRs in Exchange 2000 and 2003 , (see this link for Exchange 5.5).