Scott Kitterman wrote:
On Monday 10 December 2007 11:33, Michael Thomas wrote:
Part of the problem is that "softfail" and "hardfail" don't make
much intuitive sense. If we're going to use english terms, they
better be *very* close to the dictionary meaning otherwise they'll
be misconstrued. I, for one, was not in favor of english words for
the practices as it would force the implementor to actually read
what the draft said, rather what they could intuit from the natural
language definition.
Agreed. The larger problem is that RFC 4408 says receivers can use a Fail
result for filtering or reject the message (so it isn't clear what senders
were expecting). Having one determined set of actions for one set of
circumstances to be compliant with the spec is very useful in my experience.
I have a lot of hesitation here because advocating a particular
message disposition given a single data point is usually
a bad idea. Maybe SSP -- like gross filtering at 2821 connect
time with DNSBL's and friends -- is like that, maybe it isn't.
I'd rather let the people whose day jobs live and die by the
false positive/false negative ratio make that call. My feeling
is that if we put any advice in at all, it should be a
non-normative discussion of what the sender would prefer or
something like that.
Which isn't at all to diminish your experience. This is tough
because it's a wide ranging audience.
Mike
_______________________________________________
NOTE WELL: This list operates according to
http://mipassoc.org/dkim/ietf-list-rules.html