we need email to work a lot more often than 50% of the time.
really, one failure in ten thousand legitimate messages is
too high...we should be able to achieve one failure in a
hundred thousand messages or better.
Is that the principle why the IESG approved a mail experiment
where each and every mail with MAIL FROM != 2822-From without
2822-Sender fails, if the MAIL FROM has a v=spf1 FAIL-policy ?
No, that would be the principle that an experiment only affects
those who choose to participate in the experiment, and if the
experiment is a failure, no great harm is done to the deployed
infrastructure at large.
which of course is not true in this case.
Frank, if you were willing to write up an I-D of the form "SPF
experiment considered harmful", explaining just why this is a Bad Idea,
that might be a valuable community service.