Counting Heads

Addison Phillips [wM] aphillips at
Thu May 29 10:24:52 CEST 2003

Jeremy Carroll wrote:
 > That looks a practical problem to me.
 > Software needs to treat the normal separable use of subtags in one way
 > and these inseparable subtags in a different way.
 > We ought to have a lookup table saying which are the inseparable subtags.

I'd go further than that. Several times in this discussion we've had 
cause to refer to information about the registered tags and even some 
debate about the reasons behind a particular registration. The current 
registry contains only very minimal information, so perhaps some 
additional informative information should be included in the future.

In a new RFC one idea for registry reform might be to allow only 
concrete subtags to be registered. IOW, not "de-AT-1901". Instead you'd 
register "1901" with the note that it is designed for use with (and thus 
has practical meaning only with) "de" and its variations. If 3066bis 
were more like, say, collation (with its strength levels) this would 
make writing parsers and building exception tables much more robust.

It would also help quell debates, such as the one here, about precedent 
setting. If 3066's registry were structured this way, registering "latn" 
for use with "yi" (and later adding az, uz, and sr to it) would be 
self-limiting. The debate would only have been about whether these three 
languages should be added to the (already registered) tag. It would also 
allow productive use of the sub-tags where they make sense (e.g. we 
don't have to register a whole plethora of zh-hanx-XX tags because 
registering 'hanx' does the trick).

Just an idea.


Addison P. Phillips
Director, Globalization Architecture
webMethods, Inc.

+1 408.962.5487  mailto:aphillips at
Internationalization is an architecture. It is not a feature.

Chair, W3C I18N WG Web Services Task Force

More information about the Ietf-languages mailing list