Kari Hurtta wrote:
| o If each mail header has [RFC2047] encoded part and which encoding
| is "UTF-8", it may be a downgraded header, so decode it.
That algorithm not necessary produce original result.
What differences can arise? I can see that folding may get changed,
and whitespace may get mucked about, but is there anything else?
Changes of folding should be acceptable (and will even lead to correct
DKIM signature interpretation is the 'relaxed' canonicalization is
used).
Also header field is not necessarly downgraded header field.
In other words any valid 2321 or 2047 =?UTF-8...?= encoded word could
be encoded in the original message/utf-8 (before downgrading). Then
"upgrading" it would replace the encoded word by native UTF-8. And
that could cause havoc for header signatures. But we know this, the
issue has to be noted somewhere (maybe as "security consideration").
