Normative: Add maximize/minimize methods

[littledan]

These methods add and remove likely subtags. UTR 35 defines the relevant
algorithms in exactly the way we'd need here, so this patch just references
that specification. We'd probably want to iterate on the details of the
normative reference when integrating into the main specification.

[littledan]

cc @nciric @jackhorton

[nciric]

Thanks for the change. Looks good.

[anba]

Nit: unicode.org now supports "https", so we should use it in the links.

[anba]

And the leftover description for the no longer present [[LikelySubtags]] slot in https://tc39.github.io/proposal-intl-locale/#sec-intl.locale-internal-slots should be removed.

[anba]

Do we care about the cases where CLDR / UTR 35 canonicalization is different from IANA / RFC5646 canonicalization?

For example CLDR canonicalizes "ces" to "cs" here, which means new Intl.Locale("ces").maximize().toString() probably (?) returns "cs-Latn-CZ" here. But according to IANA "ces" is not a recognized language subtag to begin with, which in turn means new Intl.Locale("ces").toString() returns only "ces".

[littledan]

@anba Is there a UTR 35 algorithm we can reference for canonicalization?

[anba]

No idea.

[zbraniecki]

@anba - I'm confused. If CLDR canonicalizes ces to cs, new Intl.Locale("ces").toString() should already return cs.

[littledan]

We're keeping around data about things like what's the default calendar for a locale, and exposing it in Intl.DateTimeFormat.prototype.resolvedOptions(). I wonder if we should expose this in maximize() and remove it in minimize() as well.

[anba]

@zbraniecki IIUC CLDR doesn't use plain BCP 47 per RFC 5646, but instead defines Unicode Language Identifier which are based on BCP 47. Whereas ECMA-402 uses plain BCP 47 per RFC 5646, including the canonicalization defined in RFC5646 which takes only the data from the IANA language tag registry and no other sources. CLDR canonicalization seems to use the data from common/supplemental/supplementalMetadata.xml. Oh, and ICU's uloc_canonicalize is still a bit different than IANA or CLDR canonicalization, because it also attempts to recognize input from outer sources like POSIX or .NET, see http://userguide.icu-project.org/locale#TOC-Canonicalization, this is for example visible for Intl.getCanonicalLocales("es-ES-preeuro") in implementations which directly call into ICU. So, this is all a bit messy...