"hernan gonzalez" <hgonzalez(at)gmail(dot)com> writes:
> The objetionable ones IMHO are decode()/encode(), which can
> consume/produce a "non-utf8 string" (I mean, not the backend encoding)
Huh? Those deal with bytea too --- in fact, they've got nothing at
all to do with multibyte character representations. They're for
handling hex and base64 and suchlike representations of binary data.
regards, tom lane