At 02:51 PM 3/17/99 -0500, Tim Showalter wrote:
Date: Tue, 16 Mar 1999 12:29:24 -0800
From: Randall Gellens <randy(_at_)Qualcomm(_dot_)Com>
If 'discard' MUST NOT cause failure DSN generation, this could be difficult
in certain plug-in environments (such as one of the platforms where my
Sieve implementation runs).
We can either (1) make 'discard' optional, (2) say that 'discard' SHOULD
NOT cause failure DSN generation, or (3) say this is just Too Bad.
I am in favor of (3). Generating DSNs in some cases, most specifically
spam, is not a good idea.
I don't understand this.
In one particular environment where my Sieve implementation runs, 'discard'
causes the same action as if the message is undeliverable for any other
reason (a failure DSN for 'other' is generated). Why is this bad
specifically in the case of spam? It doesn't indicate the address is
valid, since there are several reasons why the server might have accepted
the message (such as being an MX for another host).
I guess this comes down to a model of what Sieve does. Randy sees it as
modifying disposition of a message; I see it as a good deal more
general.
How does a requirement for silent discard follow from purposes for Sieve
other than message disposition? Discard *is* message disposition,
regardless of any other uses for Sieve (and I use Sieve for message
annotation as well as disposition).
(1) is a very unexpected choice, but if I can't get (3), I think I like
it as much as (2). At least the behavior is consistant.
(1) is what was suggested at the Sieve BOF when I mentioned the issue. I'm
not especially fond of it.