Re: [css3-fonts] font matching of grapheme clusters

Eric Muller wrote:

> On 5.2, Cluster matching:
> 
> In 1.a, the term "completely supported" is a bit ambiguous. One
> could consider that a font without any explicit support for
> variation selectors supports all variation sequences, on the grounds
> that the variation selector is only encouraging a certain display,
> not mandating it. Given 2.a, I suspect that you intend for the font
> to provide an explicit guarantee that the variation sequence is
> honored. You could add "(including variation selectors, if any)"
> after "completely supported", if I understood your intent correctly.
> I also suspect that you actually meant something stronger, i.e. "has
> entries in the cmap for each of b, c1, c2, ..."

Agreed, that's what I was aiming for.

> Another case to consider is the joiner characters. What does it mean to
> completely support the sequence <f, ZWJ>?

> Finally, I am uneasy with the notion that "a font supports or does
> not support x". Text is not presented to fonts but to layout
> engines. It's perfectly possible to have a layout engine that
> interprets joiner characters, and regulates the application of
> ligating features accordingly. In that case, just looking at the
> font does not tell the whole story. Viewed another way, your steps
> 1.a and 1.b in combination are an attempt to describe the layout
> engine, but I don't think you can be exhaustive; if my
> implementation is more elaborate, e.g. it can deal with the sequence
> <U+00EA ê LATIN SMALL LETTER E WITH CIRCUMFLEX, U+0308 ◌̈ COMBINING
> DIAERESIS> when the font only maps e and the combining marks, do I
> become non conformant?

I'm not aiming for an exhaustive definition, but a consistent,
reasonable baseline that all user agents must support.  At a minimum,
I'd like to make it so that base characters and diacritics are
rendered from the same font if possible.  I'd also like to make it
such that a font can supply a set of diacritics and base characters
and any combination should work, rather than having layout engines
automagically canonicalize the cluster before looking to see whether
the font contains all the diacritics.

It seems to me we need some form of wording that allows wiggle room
for layout engines in the cases you list.  Any thoughts on what that
might be?

Cheers,

John Daggett

Received on Monday, 1 October 2012 05:51:58 UTC