RE: real vs. synthetic width glyphs

> From: Florian Rivoal []
> Sent: Wednesday, July 10, 2013 3:52 AM
> To:
> Subject: Re: real vs. synthetic width glyphs
> On Tue, 09 Jul 2013 20:09:25 +0200, Sylvain Galineau <>
> wrote:
> > By 'if using 1/n glyphs is not optimal' do you mean 'if *the author*
> > does not want to use the 1/n glyphs provided by the font'? If so then
> > yes, it's absolutely fine for them to be able to override the default
> > behavior.
> > It's
> > also fine for UAs to do something interesting when no such glyphs are
> > present in the specified font. I do not think any of this is really
> > the issue though.
> >
> > The argument is about *requiring* interoperable behavior when the font
> > does provides 1/n glyphs. My understanding of the resolution is that
> > it does not actually do so.
> My understanding of the resolution is the same as yours. As requiring the use of special
> glyphs when they are all available leaves the door open for Koji's
> #12 use case, I think the only question left is Elika's "MI" use case.

Ah, sorry I missed this one before replying; so I guess the answer is both Florian and Sylvain are fine not to require use of 1/n width glyphs if not all grapheme clusters have the 1/n width glyphs, so meaning this type of tweaks are totally fine to be allowed.

Elika's "MI" use case is actually much more common than mine; see the example figure 7 of the spec[1]. You'll find "10" and "27" are wider than other Kanji characters, that is, the picture is using normal proportional glyphs rather than hwid.

With some fonts, using proportional glyphs without scaling will be too wide, and may interude underlines or ruby. I agree with John, Slyvain, Florian, and Alan that hwid is better than scaling in such cases, but some fonts are perfectly fine to use normal proportional glyphs.

So the question really is, do we want to set minimum quality, or do we want to prevent some UAs doing better by additional tweaks, or both?



Received on Wednesday, 10 July 2013 14:04:43 UTC