Fixing the lost el-Latn

[Doug Ewell]

I have to admit being a bit confused about Luc's concern here.  The tag 
"el-Latn" is perfectly valid according the RFC 3066bis generative 
syntax, whether or not it is listed as a "redundant" tag or not.* 
Furthermore, unlike some of the "sgn-" tags, "el-Latn" as a generative 
tag unequivocally means the exact same thing as if it were listed as 
redundant: "Greek language in Latin script."

I agree with Luc that IANA made a mistake in not adding "el-Latn" to the 
registry, but this mistake does not really affect RFC 3066bis 
implementations at all.  Redundant entries only exist for a 
documentation purpose, to indicate what has happened to tags formerly 
registered under the old 1766/3066 rules.  The only real benefit I can 
see is for applications that support RFC 1766 and 3066 registered tags, 
but do not support RFC 3066bis, and even then this is not a matter of 
fixing the new subtag registry -- just one of retroactively fixing the 
old tag registry.

This is *not* the same as fixing the "fy" problem that occurred this 
past week, because when IANA introduced a second "fy" record, that 
actually made the registry syntactically invalid.  An application that 
reads the registry can be expected to associate one subtag of a given 
type with one or more descriptions contained within the *same record*. 
(For example, "gsw" is both "Swiss German" and "Alemanic.")  It would 
not expect to find two different "fy" records.  What if the "Added" or 
"Comments" fields were different between the two records?

It could technically be argued that IANA violated Section 5.1 by fixing 
their error, but my opinion is that leaving this clerical error in place 
would have been much worse.  And I still do not see how this is related 
to "el-Latn," since (again) that tag is legal under 3066bis whether it 
is listed as a redundant tag or not.

--
Doug Ewell
Fullerton, California, USA
http://users.adelphia.net/~dewell/

* A little pun on "redundant."


> Date: Sat, 25 Mar 2006 09:39:26 +0100
> From: Luc Pardon <lucp at skopos.be>
> Subject: Fixing the lost el-Latn (was: Re: Process Enhancement)
> To: ietf-languages at alvestrand.no
> Message-ID: <442501BE.4040003 at skopos.be>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
>
>
> Frank Ellermann wrote:
>> If they really fix the lost el-Latn in the old registry it
>> requires a (forbidden) late addition to "redundant" in the
>> new registry.
>
>    I beg to differ. I'd say it's _not_ an addition, it's the 
> correction
> of a factual error in setting up the new registry. This case is not
> provided for, so it can't be forbidden. I assume you're referring to
> RFC3066bis 3.3 where it says:
>
>    The set of redundant and grandfathered tags is permanent and 
> stable:
>    new entries in this section MUST NOT be added and existing entries
>    MUST NOT be removed
>
>    But el-Latn is not a _new_ entry, it is simply a _missing_ entry. 
> It
> should have been there from the beginning. It is missing because the 
> new
> registry was made from a wrong version of the old registry (i.e. with 
> an
> approved tag missing).
>
>    It follows that:
>
>    a) inserting a "redundant" record for el-Latn, not being a "new"
> entry, is not forbidden and hence allowed, and
>
>    b) inserting a "redundant" record for el-Latn is required because
> the new registry as it stands today is in error and must be fixed.
>
>    The error is entirely of the same nature (human mistake) and could
> (and IMHO should) be corrected in the same sensible way as the "fy"
> addition/modification mistake was corrected, i.e. by simply editing 
> the
> appropriate HTML page(s).
>
>    It's interesting to note that the "fy" case technically involved a
> deletion of an existing entry, which is expressly forbidden ("MUST 
> NOT"
> in 3066bis 5.1) and, as has been pointed out here, potentially far 
> more
> dangerous.
>
> > At some point rules are rules, and you've
>> to know how to break them without causing serious harm.
>
>    Not that I disagree, quite to the contrary, but in this particular
> case I'd say the rules were broken already by omitting el-Latn.
>
>    Specifically, RFC3066bis 2.2.8 says:
>
>       Existing IANA-registered language tags from RFC 1766 and/or RFC 
> 3066
>       maintain their validity.  These tags will be maintained in the
>       registry in records of either the "grandfathered" or "redundant"
>       type.
>
>    And in 3.3 it repeats:
>
>       The redundant and grandfathered entries together are the
>       complete list of tags registered under [RFC3066].
>
>    Currently, this is _not_ the case. As I see it, el-Latn _was_
> IANA-registered from the moment the LTR approved it. The Rule says 
> there
> must be a "redundant" record for it. There is none. The Rule is 
> broken,
> the Rule must be mended.
>
>    Unlike the "fy" case, the mending for el-Latn can be done without
> breaking any other rules.
>
>
>    Luc Pardon
>    Belgium