After doing some research...
According to section 2.1 of RFC 4646 "Tags for Identifying Languages" (http://tools.ietf.org/html/rfc4646#section-2.1), the language must be represented using the shortest ISO 639 code. If there's no ISO-639-1 (2 chars) for a language then ISO-639-2 (3 chars) code must be used (if exists). If the terminology code is different than the bibliography one, then the terminology one should be used.
Bottom line, we need to fix the Tcl API to deal with 3 chars language codes.
