Bart Schaefer a dit :
Using anything but 7-bit in headers is a calculated risk.
Using anything on internet is a calculated risk, it does fail sometimes.
Sending email *can* fail without warning.
But I think it's possible to work with the restriction that you set.
It's one thing to accept a risk to your own data, but quite another to
standardize on something that imposes that risk on others
The current draft allows both RFC2047 and UTF-8.
It can be worded so that people who are not willing to take this risk
are pushed in the RFC2047 direction.
So the people who get the risk are only those who want it.
The negative aspect is the fact some people who are unprepared for it
would receive 8 bit data they can not handle.
But they are receiving invalid 8 bit data already today.
With that wording, the draft would not increase the amount of
transmitted 8 bit data, but rather lower it, people havig a clear
reference about how to send 8 bit instead of the blank of the current
RFC1036 that leads so many to send raw local encoding, and quite a few
would prefer RFC2047 to UTF-8.