[Asrg] where the message originated

[Ian Eiloart]

--On 19 January 2009 16:45:31 -0500 Paul Russell <prussell at nd.edu> wrote:

> On 1/19/2009 6:08 AM, Ian Eiloart wrote:
>> There are at least two very good reasons for wanting to bounce rather
>> than  reject emails.
>>
>> a) when a message has two or more local recipients, you can't have
>> different policies for each recipient about what to reject. You reject
>> for  all or no recipients. There are some workarounds, but they're
>> kludgy and  cause their own problems.
>
> Perhaps that is true of your MTA, but we have been doing this with
> sendmail for several years.  This is a supported feature which can be 
enabled or
> disabled by any reasonably competent system administrator.  I would not 
describe it
> as a 'work-around' or a 'kludge', and it has not caused us any problems.

With SMTP, you have to give recipient specific defers before you see the 
body of the message. Either you defer everyone after the first recipient, 
or you split them into policy classes and defer everyone who isn't in the 
first policy class. The defers cause delays which may be severe when there 
are very many policy classes.

I guess the ideal solution would be an SMTP extension allowing MTAs to 
switch to LMTP if they're both capable. I think something similar exists.

They may even cause delivery failure. For example, I've seen an airline 
ticketing agency that failed to retry delivery of messages! However, we 
can't be expected to design around that kind of stupidity. However, if you 
defer a whole bunch of people, then reject the message for the first 
person, the sender might be forgiven for thinking that your rejection of 
the message is definitive.


>  > b) when you reject a message you can try to explain why, but often the
>  > sending MTA will throw away some or all of your explanation. Usually, 
it
>  > will wrap your explanation inside it's own. The end result is that the
>  > recipient of the bounce message (if there is one) doesn't understand 
what's
>  > going on. If you generate a bounce, you have full control of what the
>  > bounce recipient sees. For example, you could include an explanation 
of why
>  > they should lobby their domain controller to publish SPF records.
>
> Some mail servers add boilerplate text - sometimes accurate, sometime
> completely inaccurate - to every inbound DSN.  How are you going to deal 
with that
> issue? The fact is, you cannot control the operation of every mail server 
that
> might handle your completely meaningful, easily understandable, and 
perfectly
> hand-crafted 'bounce'.

Really? That's crazy. However, the ALL do it to rejection messages.

-- 
Ian Eiloart
IT Services, University of Sussex
x3148