Re: [Groff] Typesetting dashes

[Tadziu Hoffmann]

> If I understand the question properly, the implication is
> that a line might consist of words or characters at different
> point sizes, therefore the spaces separating words of the
> same point size ought to stretch proportionately.
> (Am I in the ball park here?)

Correct.

> If that's the idea, then I disagree. I can't think of a
> situation where you would want to mix point sizes on a line.

Correct.  This usually isn't done.  And therefore no harm will
be caused if the algorithm is changed.  In the overwhelming
majority of cases, the spacing will remain exactly the same
as before.

> Small caps, inferior or superior characters, and so on
> are ideally taken from fonts at the same point size as
> the regular text

Ergo, no change.

> (if they're faked by changing point size, you wouldn't want
> the spacing around them to reveal the fake).

Hmm...  I admit I haven't thought about faked small caps.
But I claim that anyone writing a fake-small-caps macro
should see to it that spaces are set at the regular size,
not the reduced size.

> Adding the same amount of stretch to each of the objects in
> the line maintains symmetry and rhythm.

I contest that.  Adding the same amount of horizontal motion
to all spaces, regardless of their natural width, changes
the relative proportions.  In the limit of large stretch, the
width of a thinspace will approach that of a regular space.
But this begs the question of why you wanted a thinspace in
the first place, instead of a regular space.

> I think the same holds true for a line with math characters
> in it that may need to change their size as part of a formula.

Well, mathmode spacing is an entirely different beast.
And the natural spacing *is* proportional to the pointsize
of the (sub)expression anyway.

> My problem with a dash joined to adjacent words with a small
> fixed space or no space is that it can easily pull those
> words together; yet in terms of readability those words
> have a weaker relationship to each other than they do to the
> rest of the words in the phrase they are part of, so pulling
> them together makes them look almost like a compound word,
> especially if the spaces on the line are stretched.

Here I agree with you 100 percent.  That's why I'd want a
stretchable space instead of a fixed space.  My choice would
be a three-quarters emdash surrounded by thinspaces (half a
normal space) (which, incidentally, will have the same width
as an em-sized dash if the regular space is nominally 1/4 em).
However, a stretchable thinspace whose width remains half that
of a normal space when the line is spread cannot currently
be had with groff.