ietf-smtp
[Top] [All Lists]

Re: SMTP traffic control

2011-10-29 22:11:17

At 19:26 +0200 on 10/29/2011, Peter J. Holzer wrote about Re: SMTP traffic control:

On 2011-10-24 15:24:49 +0100, Paul Smith wrote:
 (Of course we have bigger 'pain points' - the '10 errors and we'll drop 
 you' is the worst:
 rcpt to:<invalid(_dot_)address(_at_)bad(_dot_)domain(_dot_)com>
 4yz invalid recipient
 rcpt to:<another(_dot_)address(_at_)another(_dot_)bad.domain>
 4yz invalid recipient
 ... 8 more times
 4yz too many syntax errors, connection dropped

"Invalid recipient" should be a 5xx error, or do you expect that these
addresses will be created within the next few days?


 (now a message in the outbound queue which will never go)
 but that's off topic)

If "invalid recipient" is a 5xx error, the sender will omit the already
rejected recipients on the next delivery attempt. Eventually all invalid
recipients will be removed and - if any valid recipients are left - the
message will be delivered. (And frankly, anybody who sends a message to
tens of non-existent addresses deserves a bit of pain)

So long as at least one valid recipient has been submitted before that 10 (or whatever) bad address limit has been reached AND the rejects are done with 5XXs not 4YZs, why not just keep accepting the RCPTs? This way the message will be accepted and forwarded instead of the sender needing to remove the 10 rejected addresses and trying again (possibly getting rejected again for another 10 bad addresses). A system that tunes its rejection threshold due to having been supplied with valid addresses would be more responsive IMO than one that sends a "Go Away, clean up your list, and try again" blow-off reply. You can have a message with 25 good addresses get blown-off if 10 addresses are bad in a "Drop Dead after 10 bad addresses" implementation when if you let all of the RCPTs be accepted and responded to the message will be delivered to the good addresses when originally submitted.

If this is a mailing list message, I can see major delays occurring when some of the subscribers from a domain go away but do not unsubscribe so this type of tuning can improve acceptance and thus delivery of messages with a mixture of good and bad addresses.

<Prev in Thread] Current Thread [Next in Thread>