Characterize the context-(in)sensitivity of each setting

[hsivonen]

In reference to yesterday's ECMA-402 meeting:

For analyzing what discussed approaches makes sense for which settings, it would be helpful to characterize the different proposed settings in terms of how context-sensitive they are.

Examples:

Context-independent: Assuming that temperatures are rendered with the unit, temperature in user-preferred units makes sense regardless of the language surrounding the rendered temperature.

Ambiguity with meaning inferred from context: Decimal separator and thousand separator may overlap in different locales and short date formats may overlap with different semantics. The user may expect rendering to follow the conventions of the locale where the surrounding text originated and trying to be helpful by rendering according to the user's locale settings may be anti-helpful when the user can't tell right away if the rendering is supposed to be according to the conventions of the locale context of the site or according to the locale context of the user.

Contextual preference: I'm not sure, but to me (both theorizing on my own and looking at Windows settings), numbering system looks like a preference that might depend on the surrounding language. It seems plausible that a user might prefer ASCII digits in the context of English and Bengali digits in the context of Bangla. I'll file a separate issue requesting user stories for how numbering system preferences are used by users.

Preference is strictly scoped to the language: For collation, if the user prefers the traditional collation for Spanish, it doesn't follow that they'd want the "trad" option in the context of other languages for which a "trad" collation exists.