This function ignores any encoded-word specified in a character set not supported by the mbstring extension (e.g. Arabic windows-1256). Such encoded-words pass straight through.
RFC2047 doesn't specify the behaviour, but common agents such as Gmail, Windows Mail, etc. DO NOT ignore such encoded words. They are decoded as if they were declared to be in the local character set.
Be aware of this idiosyncrasy.
This function leaves the underscores in the string.
For example, the mime header looks like this:
$text = "=?ISO-8859-1?Q?Keld_J=F8rn_Simonsen?= <keld@example.com>";
Then I decode this string like this:
<?php
mb_internal_encoding("UTF-8");
$ret2 = mb_decode_mimeheader($text);
?>
The result is decoded string which looks OK except that the underscores between the first, middle and last names are still there.
Keld_Jørn_Simonsen
imap_utf8($text) does not show underscores, which is the correct way, except that on my Windows PC the imap_utf8() converts result to upper case, which is a known bug.
Using quoted-printable-encoding the hex-numbers must be written in uppercase letters!
Works: =?iso-8859-1?q?=3F=3F=3F?=
Fails: =?iso-8859-1?q?=3f=3f=3f?=