| Re: BACKSCATTERER problem... We are an ISP |
|
 |
|
 |
|
 |
|
 |
Author: Larry M. Smith Date: Feb 1, 2008 07:02
> On Thu, 31 Jan 2008 11:21:07 GMT Martijn Lievaart wrote:
(snip)
> | Imagine a front end mail exchanger. It can (and should!) know valid
> | users. But does it know if the user is over quota? Even if that is
> | lolved, there probably is a small race that the frond end determines the
> | user is not over quota, but by the time the mail gets delivered, the user
> | has gone over quota.
>
> That can be a real problem. One solution around that is to hold on to the
> incoming SMTP connection and make one to the next server and provide the
> RCPT info. If the user is already over quota, a good mail server will say
> so by a 4XX code rejecting the mail temporarily,
Not all sending MTAs understand the RFC1653, and there is no requirement
to implement it. So your "good mail server" rejecting RCPT TOs because
the message would put the user over quota doesn't make much sense.
> although some will do 5XX
> and make it permanent (ideally it should do 4XX for some period of time and
> go to 5XX if the issue is not resolved in a few days). The real risk is
> when the next server is accepting email and wpn't indicate that before the
> border server has completed the acceptance. One way to deal with this is
> for the border server to treat 5XX as 4XX for that user and just keep on
> trying instead of bouncing back to an uncertain sender address. After the
> period of trying, forward it to the LOCAL postmaster (not the distant one).
>
The idea that, in the event of failure, mail should be redirected to
somewhere other then the sender or the receiver, breaks SMTP.
SgtChains
| 
