Re: [css-counter-styles] status of "additional" predefined counter styles

On Thu, Sep 25, 2014 at 12:04 PM, Richard Ishida <ishida@w3.org> wrote:
> On 09/09/2014 08:48, Tab Atkins Jr. wrote:
>>
>> On Mon, Sep 8, 2014 at 8:36 AM, Jonathan Kew <jfkthame@gmail.com> wrote:
>>>
>>> In the current CSS Counter Styles draft,[1] we read that
>>>
>>> # These additional counter styles are not intended to be supported by
>>> # user-agents by default
>>>
>>> (in reference to the styles listed in the i18n WG's document.[2])
>>>
>>> However, test results[3] indicate that webkit and blink do actually
>>> implement a number of these as built-in counter styles; in particular,
>>> many
>>> of the Indian script/language names (with the exception of 'tamil'), and
>>> several more such as 'arabic-indic', 'persian', etc.
>>>
>>> A smaller number of the "additional" styles are also supported by
>>> Firefox,
>>> such as 'armenian', 'greek', and a number of CJK styles; and a handful
>>> also
>>> by IE.
>>>
>>> Should such styles be moved to the standard collection of predefined
>>> styles
>>> within the Counter Styles spec, with the expectation that all browsers
>>> should support them? If so, we need to determine which ones to move. Or
>>> should browsers refrain from implementing these "additional" styles as
>>> built-ins, so that it's clear to authors that an explicit @counter-style
>>> rule is needed in order to use them? ISTM the current situation is
>>> unhelpful
>>> from an interop point of view.
>>
>>
>> Browsers generally shouldn't be implementing random things.  However,
>> they did support a larger set than the spec currently requires; the
>> rule we used was just "what was specified in CSS2 and 2.1" to decide
>> what to keep.
>>
>> I'm fine with adding more things to the spec if browsers already
>> implement them, particularly if there are compat issues to deal with.
>> Feel free to implement what you need, and I'll reflect it in the spec
>> as necessary.
>
>
> The i18n WG discussed this in a telecon, and was also concerned that it
> might be confusing to content authors if browsers support counter-styles
> that are meant to be optional. It would be important to send out a clear
> message that you can't expect these built-in styles to be interoperable on
> all platforms.
>
> On the other hand, if we suck those into the Counter Styles spec, that may
> help, especially given that some are supported already by both Webkit and
> Blink engines, and now Gecko too (although the choice seems a little
> arbitrary). On the other other hand, IE still doesn't support the styles - I
> don't know whether they plan to support the user-defined styling - if so,
> one would hope that they will also support the additional built-in styles.
> There is also the point that older browser versions for some time won't
> support the styles it is suggested that we move into the spec, but then that
> will be an issue for user-defined styles too.
>
>
> To get some more specific information about what Jonathan was suggesting, I
> ran the tests on the latest Firefox, Safari and Chrome nightlies, and here
> is a summary of the current situation.

Thank you so much!  I was dreading having to run these tests myself,
so I'm very happy that someone else did it for me. ^_^

> Counter-styles supported by Firefox, Chrome, Opera & Safari
> (Those with * are also supported by IE, otherwise not.)  These are
> presumably the ones we are talking about adding to the spec.
>
> arabic-indic
> persian
> armenian*
> lower-armenian
> upper-armenian
> bengali
> devanagari
> georgian*
> lower-greek*
> gujarati
> gurmukhi
> hebrew
> cjk-earthly-branch
> cjk-heavenly-stem
> hiragana
> katakana
> hiragana-iroha
> katakana-iroha
> kannada
> khmer
> cambodian
> lao
> malayalam
> mongolian
> myanmar
> oriya
> telugu
> thai
> tibetan

Yes, we should add all of these, since they're supported.

> Additional counter-styles supported by Firefox, but not others
>
> cjk-decimal
> korean-hangul-formal
> tamil

The first two are already in the spec.  Tamil is a maybe.

> Counter-styles supported by Chrome, Safari & Opera (with small
> inconsistencies from spec) and not supported by Firefox. If Firefox was to
> add these quickly, they may qualify for inclusion in the spec, however the
> small divergences in the current implementations compared to the spec may
> need some attention. For a list of divergences, see the notes at
> http://www.w3.org/International/tests/repository/predefined-counter-styles/results-predefined-counter-styles#ethiopic
>
> afar
> oromo
> sidama
> tigre

Like tamil, these are maybes.  We can possibly just drop them, or else
FF can implement them and I can pull them into the spec.

~TJ

Received on Thursday, 25 September 2014 19:18:13 UTC