I agree with all your definitions except this one. I would say that
softfail is closer to neutral than fail. The idea is that, if
possible, you should give some sort of warning to the user that they
are using a discouraged/depreciated IP address.
For example, you could give a graylisting-like SMTP 4xx temp-fail with
a warning message the first time the an email shows up that softfails,
but accept it the second time. Or, an MTA could display a warning if
it notices a Received-SPF: header that shows a softfail. Or,
something could be added to the Subject: line or even the message
body.
I see. So fail, softfail, and neutral all pretty much mean the same
thing as far as the basic structure of the policy, but they are in
descending order of severity. softfail (and perhaps neutral to a lesser
extent) can be used as hints in an overall anti-spam policy, but should
not be used to block indiscriminantly. That about right?