Rasmus and I talked about this some more yesterday, and I think there is
an alternate, better approach.
[...snip...]
Love it, but we need a few fairly big warnings: "When unicode_semantics is
enabled, the $_SERVER['UNICODE_HTTP_INPUT_FAILURE'] variable MUST be checked
blah blah blah" put into the manual.
Partial decoding of the request body/string should go a long way towards
easing the issue for those 20% cases where input decoding will just fail.
-Sara
P.S. - Has anyone considered offering up an RFC to IETF or W3C about adding
a header to the spec? Or just asking the nice Firefox folks to blaze the
trail with an X-header? PHP can't be the only web-language dealing with
this issue.