Re: [Ietf-languages] Adding prefixes with dialect variants to Occitan orthographic variants

[Mark Davis ☕️ <[e-mail]>]

See message in other thread, also




Mark


On Mon, Apr 19, 2021 at 12:49 PM Doug Ewell <[e-mail]> wrote:

> Mark Davis wrote:
>
> > However, since that time, truncation is far less of a concern, and the
> > bigger issue is that BCP47 doesn't provide a well-defined canonical
> > order of all prefixes. Without that, comparison of language tags is
> > fundamentally flawed.
>
> I agree with Addison that an ordering mechanism does exist, between the
> text of BCP 47 (Section 4.1, item 6) and the Prefix values in the Registry.
> There may be some gaps, making the mechanism less than perfect
> mathematically.
>

Sorting algorithms, for example, kinda depend on mathematical correctness.


>
> > So CLDR extends BCP47 to simply order the variants alphabetically. Any
> > application that cares about variants (and there are very few) can
> > treat those variants they care about simply as a set. Thus both
> > oc-gascon-phonipa and oc-phonipa-gascon are both interpreted by CLDR
> > as [language=oc, variantSet={gascon, phonipa}].
>
> However, processes that use BCP 47 but not CLDR need to work with what is
> in BCP 47, and that includes adding Prefix values to guide the use of
> multiple variants.
>

A general-purpose implementation that depends on an ordering of variants
that is not *required* by BCP47 is a fragile implementation. And note that
the canonicalization process in BCP47 does not put the variant tags in a
particular order. The incoming language tags sl-rozaj-biske and
sl-biske-rozaj are both equally valid.

>
> >> on which combinations might be valid
> >
> > I know what you mean there, but this would be better phrased as "on
> > which combinations are useful". All combinations of registered variant
> > subtags are valid; it is just that some of them are pointless.
>
> Yes, I was imprecise in my use of "valid", which is a term of art and
> needs to be used precisely. I meant combinations that would be "meaningful"
> in Occitan or have a snowball's chance of existing in non-artificial text
> samples, not "allowable" by the rules of BCP 47 or any other specification.
>
> --
> Doug Ewell, CC, ALB | Lakewood, CO, US | ewellic.org
>
>
>