Re: [groff] 03/09: tmac/an-old.tmac: Stop remapping ` and '.

[Ingo Schwarze]

Hi John,

this seems a bold, novel, and interesting idea to me.  I'm not yet
completely sure whether it's viable, but it might be.  I'm not sure
how difficult it would be to implement in groff.  In mandoc, it is
certainly possible.

The rest of this reply addresses a few details.

Yours,
  Ingo


John Gardner wrote on Mon, Nov 02, 2020 at 11:24:49PM +1100:

> I think it's time we took constant-width fonts seriously when
> preparing output for the terminal. Here me out on this:
> 
> You see, I've never, *ever* needed fancy typography when typesetting
> code samples, which are invariably set in a fixed-pitch (monspace)
> typeface.

I agree.  That's often done, it's not always done, but would almost
always be the right thing to do.

> The use of a fixed-pitch font stipulates the author's expectation
> that that verbatim input should *stay* verbatim.

This may not be true in every conceivable typesetting situation.
But in manual pages, i do agree that's a very reasonable assumption.

Besides, even in the very unusual case of an author desiring
curly quotes in a fixed-width font in a manual page, they can
readily use \(oq...\(cq even when the rule you propose is in force.

> On the other hand, nobody in their right mind would deliberately
> write source-code in a proportional typeface, so we can trust the
> use of Georgia or Helvetica to indicate body text - stuff
> for reading and styling.

There might be exceptions in some unusual typesetting situations
(e.g. obfuscated code contests / ASCII art and the like), but in
manual pages, again, i think this is a completely reasonable
assumption.

Besides, even in the very unusual case of an author desiring
to display code in a proportional font, they can readily
use \(aq and \(ga to get ASCII ' and ` printed.

> So in all seriousness, we should be revising our man pages to produce
> better-looking monospaced output. Have you tried viewing your average
> man(7) user's "EXAMPLES" section as PostScript? It's rarely pretty
> because authors don't care about switching between proportional and
> constant-width typefaces. After all, why *would* you if you only care
> about terminal output?

While your argument makes sense, note that part of the infrastructure
already *is* in place:

 * In man(7) - or more precisely, in man-ext - the .EX macro
   already does .do fam C .
 * In mdoc(7), .Bd -literal .Dl .Li .Ql .Dv .Er .Ev
   already do .nop \*[doc-Li-font]\c
   which is \f[C] in troff mode and \f[R] in nroff mode

> Currently, \f(CW is meaningless in -Tutf8 or -Tascii output, but it could
> very well be exploited to know which quotes are appropriate to remap, and
> what portions of an author's document to keep our grubby meathooks off of.

>From a user's perspective, that does make sense to me.

A different way to talk about it, maybe more clearly, would be to talk
about a "literal mode", which would have two distinct effects:

 * set ASCII characters verbatim, even ' - ^ ` ~
   with the only exception of \ which still needs to be \e or \[rs]
 * use a fixed-width font

What do you think?
Is that logically sound, and is it feasible?
In mandoc, it is clearly feasible.

It also seems manageable from the perspective of maintaining large
collections of manual pages, and it will definitely improve average
typographic quality.