On Sun, Apr 30, 2006 at 09:03:03AM -0700, Mike Bird wrote: > I think you're following the logic already followed by many ISPs. > The next step is to consider what should happen after the message > has been in the recipient's ISP's mail queue for a few days. I think this is the worst possible way to handle full mailboxes. I think that if the message is accepted for delivery to local mailbox, it is better to bite the bullet and just deliver it. If I send someone email, I would expect that if the servers are up and they accept the mail, it does not silently sleep in the queue for days. If there are warning bounces and a rejection bounce, the backscatter problem keeps on multiplying and if there are no warnings, the sender will falsely believe that the message has been delivered. I think bouncing immediately is better than to wait, warn, warn again and then bounce. But the best resolution is to just deliver it at update the front servers to temporarily reject further messages. > Bounces should be minimized. In many circumstances, they cannot > be avoided. True. Absolutely. > SORBS is the only well-known RBL which lists IPs for backscatter > as a result of SORBS' own honeypot addresses being compromised. It might be a good idea for other lists to start doing also.
Description: Digital signature