Logged In: YES
user_id=79902

That's exactly the way it is defined to work, using just the
low byte from each character. The [encoding convertto] just
transforms the characters into a sequence of bytes (with
each resulting byte encoded in a single character in the
range \u0000-\u00ff) which are then handled in the way you
describe.

The key point is that Tcl is using a different
interpretation of bytes and (especially) characters to what
you expected.

Keeping bug open while I consider if a documentation update
is needed.