Re: [math-on-web] CG meeting minutes, 2018/01/18

> 
>> - is the height of the container that needs to be surround by the fence, within certain margins?
>> - if so, use a single glyph representing the fence: U+007d "{"
>> - if the height is above those bounds, create a "stacked" fenced, by putting a glyph at the top U+23a7 "⎧", a glyph at the bottom U+23a9 "⎩", a glyph in the middle U+23a8 "⎨" and for the rest of it, a "repeating" glyph, U+23aa "⎪" (and so on for all the different types of fences)
> Before choosing a way to proceed, I tried to understand how TeX works. There is several prebuilt sizes of each fence (3 ou 4). TeX first tries to use these sizes and if it needs hiegher, TeX builts the "stacked" fenced as you discribed. But what about italic and other orientations ! ?

“Stacking” is also performed for horizontal fences in the same way. Fences in traditional mathematical typography are not displayed differently in italic. See [1] for example.

> So,I wonder why nobody tried to sent a simple font file with the right sized glyph draw to the glyph engine (Harfbuzz or other) as TeX but for each sizes and not only for the first four.

Because the notion of “font size” for a glyph engine usually is centered around the “em box”, which glyph of a font are drawn (somewhat) into. It doesn’t say anything about drawing a glyph so that the distance between the first and the last pixel of a glyph is a specific value, which is what you would want for stretched ascenders.

> I think that drawing a bracket as a curve is simpliest and economics than the algorithm that puts glyphs over other glyphs.

Indeed, it is simpler, however it does not follow the traditions of mathematical typography. In particular, consider that fences for a latin sans-serif font look distinct from a serif font. So, for the best possibly mathematical typography, it is better to use the glyph information provided by the font file. Note that OpenType math fonts include a MathVariants table that specify how to correctly draw stretched glyphs. [2] It would be advantageous if the HTML/CSS layout engines made use of this information when available.


[1] http://www.shearsoneditorial.com/2012/07/typographical-conventions-for-mathematics/ <http://www.shearsoneditorial.com/2012/07/typographical-conventions-for-mathematics/>
[2] https://www.microsoft.com/typography/otspec/math.htm <https://www.microsoft.com/typography/otspec/math.htm>