Specifically, end-nodes typically extract data from MIME and use it within
the local context of that application *ONLY*. The encoded data is RARELY
used inside of other applications, and the extracted data is typically
used for this purpose, when needed. On the other hand, domain names are
frequently cross-populated among applications, and as the argument
regarding clipboards shows, the encoded form will be the norm.
Many people copy-and-paste MIME content all the time, just as they
will copy-and-paste IDNs. Non-mail programs look at mail mailboxes
for things like harvesting for address books or PIMs, and those
programs interpret the MIME because they store the text in some
internal format that is certainly not MIME.