Re: [css3-fonts] font selection for Unicode Variation Selector

On 2/26/2011 12:32 PM, John Hudson wrote:
> [bcc'd to mpeg-OTspec]
>
>
> Asmus Freytag wrote:
>
>> In this context, the way a CSS statement like
>>
>>  font-family: 'My Favorite Font', 'Some generic font', 'Special IVS 
>> Font', 'Generic IVS Font';
>>
>> ought to work is that it sets up a composite font selection on the fly.
>
>> Each grapheme cluster would then be rendered at the first font 
>> encountered that is able to render it.
>
> But using an actual composite font, perhaps it could be possible to 
> define which grapheme clusters are rendered by which component fonts, 
> avoiding the problem that I understood Koji-san to be be describing, 
> in which the preferred rendering of individual grapheme clusters might 
> not be capturable by fallback ordering. Taking your font-family order, 
> for example, what if both 'Special IVS Font' and 'Generic IVS Font' 
> were able to render grapheme clusters foo and bar, but the preferred 
> rendering for foo is that provided by 'Special IVS Font' and the 
> preferred rendering for bar is that provided by 'Generic IVS Font'.

Correct. A simple fallback scheme can only ever pick the first font 
that's able to render the entity.

The composite fonts that I implemented in my work allow you to do either 
a font cascade, or, if you provide the extra information, reserve a 
range for a particular font.

The former has advantages when you don't control which fonts are 
actually available, the latter is the method of choice when you describe 
the interaction between a known set of fonts that are guaranteed to be 
there.

The combination of both lets you describe which fonts to use from your 
favorite set, but allow a generic font (or set of fonts) to act as backup.

Finally, beyond that, you can allow the operating system to provide 
further fallbacks.

>
> At the moment, if I recall correctly, the proposed composite font 
> format only specifies which Unicode ranges would be rendered by which 
> component font, but it seems to me that it might be desirable to be 
> able to specify individual variation selector sequences as overrides.

Right. I haven't had to deal with clusters in my work - they pose their 
own challenges, but for things like IVS and to some extent combining 
marks, you do need to take them into account.

Received on Saturday, 26 February 2011 21:29:01 UTC