Opentype and optical sizes

Simon Cozens

Since this has come up in a few threads, I wasn't sure where to ask, so I'll start a new thread: Was there a reason why OpenType didn't allow for adjusting the design at different optical sizes? Of course you can (and sometimes do) have multiple fonts file cut for different target sizes, but the whole point of digital fonts is that you can set them at any size you like. Rather than having Foobar 12 and Foobar 18, wouldn't it make sense to have a Foobar which interpolated between the two designs at the appropriate sizes? Did GX do any better? And is this on the cards for future OpenType?

What reminded me to ask was that Arne's font here has lovely inktraps which do the right thing at text sizes but look way too conspicuous at display size. A font system which allowed for scaling between two different designs would allow this to work nicely at both sizes.

James Puckett

The size tag does this. As for interpolating on the fly, that might work for Latin, but users of complex scripts might not want to wait.

Simon Cozens

OK, I did not know about that tag. Basic research suggests that next to nothing supports it, which is a shame. :-(

Khaled Hosny

XeTeX supports size feature, and HarfBuzz has the API to read it. There is also the new optical size stuff in OS/2 that no one (even MS, it seems) knows how it can be used for font selection.

John Hudson

The <size> feature was mostly a non-starter: it's a hack of the GPOS data structure to store information to make optical master font selection within a family. It never surprised me that most software developers ignored it.

The OS/2 size values are only half of a solution: they require a name convention to identify fonts that differ only by size.

With regard to interpolation: there's been a lot of talk about this over the past year, e.g. at the TYPO Labs event in Berlin in May, and my suspicion is that we'll see something concrete within a year. Webfonts have resurrected interest in interpolation and extrapolation font formats.