Framing the Alpha-3 turf

John Cowan cowan at
Thu Dec 21 15:55:34 CET 2006

Don Osborn scripsit:

> I've waded through part of RFC-4646 but am not sure that it addresses this
> kind of issue. Thanks in advance.

This is what I believe the final ISO structure to be (note that I don't
speak with anyone's authority but my own).  I hope it answers all your

639-3 will be a comprehensive list of individual languages and
macrolanguages, with mappings between the two groups.  Each language
and macrolanguage will be assigned a unique 3-alpha code element.

639-5 will be a list, but not a comprehensive one, of language
collections, mostly genetic groupings but also some other types.  Each
group will be assigned a unique 3-alpha code element, not overlapping
with the ones in 639-3.  The two RAs will cooperate through the RA/JAC
to ensure this.

639-2 will be the union of subsets of 639-3 and 639-5, providing a
list of code elements known to be useful for institutional purposes.
639-2 also provides a small fixed set of alternative 3-alpha code elements
for backward compatibility, distinct from those in 639-3 and 639-5,
for certain languages (the so-called "bibliographic" code elements).
639-2 will not assign any new 3-alpha code elements of its own after 639-3
and 639-5 are fully up and running; however, code elements from 639-3
and 639-5 may be added to 639-2 if they meet its criteria for inclusion.

639-1 will provide 2-alpha code elements for a fixed subset of 639-2
entries, representing the most commonly used languages.  Systems using 639
code elements may use 2-alpha code elements exclusively (in which most
languages and collections cannot be represented at all), 3-alpha code
elements exclusively, or the rule "2-alpha where possible, 3-alpha where
necessary", which is what RFC 4646 does.  One advantage of the mixed rule
is that in every case where 639-2 assigns a bibliographic alternative
code element, 639-1 also assigns a 2-letter code element, eliminating
the necessity of choosing between bibliographic and 639-3-compatible
code elements.

639-6 will contain a comprehensive hierarchy of genetic groupings,
languages, and language varieties.  Wherever a 3-alpha code element
from 639-3 or 639-5 is available, it will be used; in all other cases,
4-alpha code elements assigned by the 639-6/RA will be used.

(639-4 is expected to explain all this in detail; it will contain no
code elements.)

John Cowan      cowan at
        "Not to know The Smiths is not to know K.X.U."  --K.X.U.

More information about the Ietf-languages mailing list