On Wed, 2006-12-06 at 10:38 +0000, Chris Lightfoot wrote: > If you're going to start announcing temporary errors as
> permanent, it's hardly a surprise that people are going to
> start treating announced permanent errors as temporary!
> There isn't much machine-readable information in the SMTP
> error codes, and it hardly seems wise to reduce the amount
> further by using them inappropriately.
But this is the whole point: the MTA has no knowledge of *why* the user
is over quota, just that they are. And that means the message cannot be
delivered (presumably after a configurable period), this it's a
permanent error.
In most cases with quotas (in my experience) a frontend MX accepts the
message and then handles the intervening 4xx errors from the "postbox"
server until such time as its' retry times are exceeded. Then, and only
then, is a 5xx sent back.
In the case of a machine which is both MX and postbox, the receiving MTA
would usually return a 4xx to the sending MTA - which will then generate
a 5xx itself after its' retry times are exceeded.
Quota errors are not in themselves SMTP errors in any way, they're
system errors. How the MTA admins decide to handle them is entirely up
to them, but at some point a permanent SMTP failure should (must?) be
created and returned to the sender if the message is undeliverable.
I for one do not want to hold on to over-quota messages for longer than
is necessary, but it's my decision as the admin of the receiving system
to create and administer the appropriate policy, not the sender.
