Message-Id: <6.2.3.4.2.20050828234319.0597ed50@mail.afrac.org>
Date: Mon, 29 Aug 2005 02:22:49 +0200
To: "Peter Constable" <petercon@microsoft.com>,
	"LTRU Working Group" <ltru@ietf.org>
From: "JFC (Jefsey) Morfin" <jefsey@jefsey.com>
In-Reply-To: <F8ACB1B494D9734783AAB114D0CE68FE06EDF08E@RED-MSG-52.redmon
	d.corp.microsoft.com>
References: <F8ACB1B494D9734783AAB114D0CE68FE06EDF08E@RED-MSG-52.redmond.corp.microsoft.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Cc: iesg@iesg.org, ietf@ietf.org
Subject: STD (was: Last Call: 'Tags for Identifying Languages'toBCP)
Precedence: list
Sender: ietf-bounces@ietf.org
Errors-To: ietf-bounces@ietf.org

An exchange on WG-ltru documents (I do not say "support": the reader 
will judge) the positions I support.

It involves:
- Peter Constable: one of the initiator of the project and author of 
ISO 639-3 which lists 7500 languages and is used in building langtags
- Doug Ewel: author of the Draft concerning the initial content of the registry
- Debbie Garside: the author of ISO 639-6

At 22:20 28/08/2005, Peter Constable wrote:
>[I'll preface this reply by saying that we don't want to spend too much
>time discussing issues that are not of immediate concern while we've got
>the matching draft and IETF last call on the registry drafts to deal
>with. So, I won't pursue this thread much longer.]

The proposed Draft is based upon ISO 639-1,2,3 lists of language 
names. ISO 639-6 is a list of language use names and IDs. The 
proposed langtag is an arbitrary limited compound of three 
information: language name, script and country. A language 
identification MAY call for far more elements, and deliver much more 
information. However these three basic elements are necessary to sell 
lingually related products  (contract, ads, documentation, bills) and 
identify the current status of the art "locales" (CLDR project).

The alternative seems to be:
- GO for an e-commercial only multilingual internet, for ever.
- NO we do not want the Multilingual Internet to be only commercial.
The decision is NOW. And we understand Peter and the authors wants to 
win now, because they have real needs to address now.

But I do not think there is a need for anyone to "win". There is a 
third response.
- GO for an e-commercial multilingual internet support now, as 
default/immediate solution
- YES to a generalised Multilingual Internet hooked to the RFC 3066 
Draft how poor it is, using its reserved ABNF hooks.

This means that:
- "fr-Latn-fr" is the default tag based upon ISO 639-1/2/3
- "x-fran" is a private use tag based upon ISO 639-6
- "0-jefsey.com:franver" is my vision of the French at the Palace of 
Versailles. Documented by an ISO 11179 conformant system (see below)

> > From:Doug Ewell
> > I'm a bit surprised that a work characterized as a work-in-progress
> > and not yet ready for public review is nevertheless deemed ready
> > to be considered as a draft international standard.
>
>Debbie at no point said that it was -- and it is not. It will be
>December at the earliest that it can be registered as a CD, and it must
>successfully complete a three-month ballot as CD before it can be
>registered as a Draft International Standard. So last spring of 2006 at
>the earliest.

This means that this debate is only to lock a _final_ ABNF via an 
accepted RFC and a loaded operationalIANA registry _before_ a simpler 
solution is available three months from now....

> > > In other words, in the system as proposed, you could
> > > use either the alpha-4 representation or the unique DI to find the
> > > closest 639-1,-2,-3 or -5 tags should you so wish.
> >
> > But in language tags, either one value needs to be canonical -- sorry,
> > "preferred" -- over the others, or else the duplicative values should
> > not be added at all.
>
>Your statement doesn't contradict anything that Debbie has said,
>provided the context is ISO 639-6 alone. If we were to talk about
>incorporation of ISO 639-6 into a revision of RFC 3066, however, then
>duplication would become an issue for consideration.

This is the WG-ltru Charter that all the ISO codes be included. As a 
user I am not much interested in mixing four formats only to please 
Peter Constable and/or Debbie Garside. All the more than the issue is 
the addition of the script information to document ... oral 
expression and they miss computer(ised?) languages (definition?) and 
all this is through computers.

>For clarification of Debbie's statement, in the model of ISO 11179, we
>have metadata elements that consist of a data element concept, such as
>'English', and a representation for that, such as "en" or "eng" (these
>would be distinct representations belonging to different value domains).
>Within an metadata registry, a registry item corresponding to 'English'
>can have a Data Identifier (DI), which is a unique identifier *within
>the registry* for that administered item; in this example, that DI could
>be any number of strings, though "English" would be among the better
>choices.

Nice to see that ISO 11179 is accepted now. Peter Constable and the 
WG-ltru have opposed the reference to ISO 11179 model. This model 
permits to conceptualise languages and to include in their 
description an unlimited number of additional elements. Roughly it 
means that ISO 639-3 is a table of codes (names) related to non 
documented languages. While ISO 639-6 wants to be a root to a base of 
objects describing languages.

The Draft proposes a very limited version of that base with three 
columns only. This is enough in many cases. But not in an increasing 
number of cases. Hence the possibility to use the Draft as a default. 
Since the three elements of the Draft's langtag are also in the 
language object base. CLDR (Unicode locale project) is a langtag 
related base. But ISO 11179 totally open the concept, like C 
structures: data they may indefinitely expand as metadata. Each data 
has  the possibility to become a metadata describing a new dataset.

For example "fr-latn-FR" can use the three codes as data. This is the 
case of the Draft.

Or it can use  the ISO 639-1 "fr" data, as the meta data describing 
all the French dialects and many new names to be listed (which are in 
ISO 639-6). "Latn" element of ISO 15924 for Latin script can become a 
meta data introducing French charset within a (to be defined) Latn 
concept, and also include elements such as founts, thickness, color, 
etc. etc. This is the approach I support: this is simple enough in 
extending the DI concept in a network, but this has be to discussed 
within ISO/TC 32 (or in a dedicated WG on reference centers?)

> > Will Linguasphere provide the mapping between the new alpha-4 codes
> > and ISO 639-1/2/3, or is that something a group like this would have to
>do?
>They will be (and must) providing that.

This insures full compatibility between all the visions using ISO 
639-X codes (documented by the ISO 639-4 guidelines under work).

> > I agree that the broad question of "what is a language" is out of our
> > scope. The more specific question "what is a taggable language
> > distinction" is perhaps more germane.
>
>Not an unreasonable suggestion.

This is a major step ahead for common understanding!
My request, plea, begging, for a definition of what we intend to mean 
as a "language" in the Draft contest, might after all be listen to.

I would then advise that the Draft is sent back to the WG-ltru, with 
the suggestion that a lexicon is provided which would define what is 
a "language", a "script", a "country", and the purpose (informative, 
descriptive, normative?) of a langtag. This might be a big step ahead.

jfc



_______________________________________________
Ietf mailing list
Ietf@ietf.org
https://www1.ietf.org/mailman/listinfo/ietf