Re: Increase default `line-spacing' to 0.05, 0.10 or 0.15 [proposal]

[Augusto Stoffel]

On Thu,  6 May 2021 at 13:10, Eli Zaretskii <eliz@gnu.org> wrote:

>> From: Augusto Stoffel <arstoffel@gmail.com>
>> Date: Thu, 06 May 2021 11:26:45 +0200
>> Cc: Jim Porter <jporterbugs@gmail.com>, emacs-devel@gnu.org
>> 
>> Comparing the picture in [1] and what I see if I put the block cursor
>> over an "H", I conclude that monospaced fonts come with some built-in
>> leading.  In other words, they are "bastard fonts" [2], and a nil for
>> `line-spacing' doesn't mean "solid" typesetting.
>
> I don't think this is true.  First, what you cite refers to so-called
> "manual typesetting", not to digital typography.  And second, the
> monospaced fonts we use define ascent and descent for each glyph, and
> we use that and nothing else (unless the font is broken, in which case
> we have fallbacks in place, but that's not relevant to the issue at
> hand).  AFAIU, the ascent and descent of each font glyph in a
> monospaced font is set up such that they accommodate both the tallest
> glyph and the lowest one.  Perhaps that is why you think they have
> some "leading", because you probably didn't examine all of the font's
> glyphs (we generally use as default fonts that cover many popular
> scripts).

Okay, we could ask a font designer if this standard glyph size is chosen
so as to provide a reasonable default leading (as I was assuming), or
merely to fit the ascent and descent of the tallest/deepest glyphs.  But
I would bet that the answer is both, since designing a font, especially
monospaced, involves lots of compromises.

Anyway, I think we can agree that setting `line-spacing' to nil is *not*
analogous setting \baselineskip=0pt in TeX.  That's all I wanted to
say there.

>> Kerning, tracking and other forms of microtypography are case-by-case
>> adjustments, intended to be basically subliminal, so I'd say yes, the
>> overall condensedness is pretty much a fixed characteristic of the font.
>> And these things are out of reach for Emacs anyway.
>
> They are not out of reach for us, because the shaping engine(s) we use
> know how to apply these features.  We just don't use them by default
> with most characters, that's all: we don't ask the text-shaping engine
> how to render each sequence of characters we need to display, we only
> ask it about some relatively rare sequences.  The reason for that is
> that under the current design of the display engine, calling the
> shaping engine is very expensive, as it requires calling out to Lisp,
> which then calls back into C.  So we only do that when necessary.  But
> a Lisp program can change that.

You are right, I should have said "these things [such as kerning] are
not relevant for monospaced fonts".