Addison Phillips [wM]
aphillips at webmethods.com
Mon Nov 15 19:40:21 CET 2004
The recommended prefix is informative, not normative (just as its name suggests). It accomplishes the same task as registering whole tags, without restricting variants from use in the generative mechanism. This has the side effects that you note.
Both 'ar-scouse' and 'en-boont-scouse' are well-formed and valid. The former tag has informative information that indicates that it would be unwise to form that tag. The latter tag doesn't have that information in the machine-readable part of the registry (one would have to look at the actual registration to know it). In practice, both tags violate "rule 0" of language tags: "Choose tags wisely."
In authoring the spec, we didn't solve every possible tag formation problem in a normative manner. The tags you give in your email below are preposterous, if "valid", but no more so than 'zh-CO' (well-formed and valid under RFC 3066) or 'en-Egyp-CH' (ditto for draft-langtags). Writing a registry file format that would describe the valid ranges and exclusions for registered variants can probably be done, but at the cost of greater complexity (and further, unnecessary, delay). I don't see that we need to solve this problem for the informative information in an "RP", except via usage guidelines.
Addison P. Phillips
Director, Globalization Architecture
Chair, W3C Internationalization Working Group
Internationalization is an architecture.
It is not a feature.
> -----Original Message-----
> From: Doug Ewell [mailto:dewell at adelphia.net]
> Sent: 2004年11月13日 20:49
> To: ietf-languages at iana.org
> Cc: Addison Phillips [wM]; Mark Davis
> Subject: en-boont-scouse?
> Variant subtags under RFC 3066bis are defined as having a "recommended
> prefix," such that they are considered appropriate for use with language
> tags that match that prefix and not appropriate (though still valid)
> with those that do not.
> For example, the variant subtag 'scouse' has a recommended prefix of
> 'en', so that "en-scouse" is sensible and valid, while "ar-scouse" is
> silly but still valid.
> Variant subtags can have more than one recommended prefix (hereafter
> "RP"), although no such subtags are currently proposed. Also, a
> language tag can have more than one variant subtag. For example, a
> while back, Han Steenwijk proposed subtags for sub-dialects of the
> Resian dialect of Slovenian. In the RFC 3066bis environment, these
> would be variant subtags with an RP of 'sl-rozaj' instead of merely
> 'sl'. This way, for example, a user could write "sl-rozaj-ravan" to
> indicate the Ravanca sub-dialect. Switching the subtags around, as
> 'sl-ravan-rozaj', or would still be technically valid but would no
> longer follow the intent of the RP mechanism, because the qualifying
> prefix for 'ravan' would be 'sl' and not 'sl-rozaj'.
> Variant subtags can be registered to represent concepts other than
> dialects and sub-dialects. The subtags '1901' and '1996' are intended
> to be used with the prefix 'de' to indicate the spelling conventions
> that were first promulgated in those years (more or less). If there
> were *also* a variant subtag to indicate a specific sub-dialect of
> German, say 'xyzzy', it would be perfectly reasonable to write any of
> the following:
> and also:
> Of the latter four, the first and third would somehow have to be
> considered equivalent, as would the second and fourth.
> The point is that 'xyzzy' would only have to have an RP of 'de', and not
> 'de-1901', in order to be used reasonably with 'de-1901'. It would
> clearly be appropriate to use these variant subtags together, as well as
> separately, because they do not represent mutually exclusive concepts.
> I am wondering about the "recommended-ness" of a language tag that
> contains two or more variant subtags that have the same RP, but which
> are *not* intended to be used together, because they do represent
> mutually exclusive concepts. Currently, the following language tags
> would be not only valid, but would also adhere to the letter of the law
> concerning RPs; however, they are obviously preposterous:
> zh-guoyu-hakka-xiang (or pick any two)
> Switching the order of the variant subtags in the tags above would be
> equally preposterous.
> What is the policy, or what should it be, concerning such tags? It
> would not be desirable to restrict the RP concept so that any variant
> subtags in the "base" tag must be indicated in the RP, because that
> would remove legitimacy from the "de-1901-xyzzy" example above. Yet
> "en-boont-scouse" (which passes the RP test) is clearly as silly a tag
> as "ar-scouse" (which fails it), and for the same reason.
> Please note that I am asking this for my own clarification, NOT to stall
> or derail the approval process for RFC 3066bis in any way. My validator
> says that "ar-scouse" violates the RP test, but "en-boont-scouse" does
> not, and I want to know if that is correct.
> -Doug Ewell
> Fullerton, California
More information about the Ietf-languages