Hi,
after reviewing all feedback, i come to the conclusion that there is no
consensus for the change:
OPPOSED: Mike Bianchi, Tadziu Hoffmann
SCEPTICAL: Ralph Corderoy
NO EXPLICIT PREFERENCE STATED: Doug McIlroy, Werner Lemberg
APPEAR TO NOT OBJECT: Dave Kemper, Jason McIntyre
IN FAVOUR: Anthony Bentley, Bertrand Garrigues, Colin Watson,
Ingo Schwarze, Jeff Conrad, Ted Unangst
I think that established behaviour can be changed if there is
consensus. I also think that established behaviour can be changed
if there is an overwhelming majority together with strong technical
reasons or very significant benefit from the change. I doubt that
it is wise to change established behaviour based on a majority when
the benefit is, as John aptly put it, "nice to have" and substantial
arguments have been presented for both sides that nobody managed
to fully reconcile.
So i think it is best that i do not commit my patch, do not change
the behaviour of the mandoc toolkit either, close the bugtracker
ticket as WONTFIX in a few days, and instead consider whether the
mandoc_char(7) manual or the "ASCII Output" section in the mandoc(1)
manual should instead explain why the ASCII output device renders
single quotes as 0x60/0x27 - which, as the thread has extensively
demonstrated, is no longer obvious even for some experienced computer
users nowadays.
Sorry for the considerable time you all spent on this; while i
expected that some concerns i overlooked might possibly be raised
(that's why i brought it up rather than just committing it), i did
not quite expect that such a large number of different aspects might
need to be taken into account for such a deceptively simple-looking
matter.
Yours,
Ingo
P.S.
My final summary of the main arguments, for easy later reference:
BENEFITS WOULD HAVE BEEN:
-------------------------
- stop relying on a meaning of ASCII 0x60/0x27
that conflicts with ISO646 and with Unicode
- compatibility with modern (Unicode-compatible) fonts
that treat ASCII 0x60 unambiguously as "accent grave"
(admittedly, people often use -Tutf8 together with those)
- but still useful when using LC_CTYPE=C temporarily, for example
in build system contexts or when logged into remote machines,
and for those people always using LC_CTYPE=C with modern fonts
- symmetry with the ASCII rendering of \(lq
- symmetry with groff error messages etc.
- compatibility with the GNU coding standards
https://www.gnu.org/prep/standards/html_node/Quote-Characters.html
(admittedly, for roff output, historic precendent may be more
important than GNU coding standards, so i give this argument last)
DOWNSIDES WOULD HAVE BEEN:
--------------------------
- ASCII clearly states that 0x60/0x27 are intended to be used
as symmetrical single quotes (even though it also permits the
use of the same bytes as accents)
- several people still use fonts that support this usage of ASCII,
so for them, the output would actually look worse
- it would be a change of established behaviour
for a benefit of the "nice to have" class
- may cause issues when post-processing -Tascii output with scripts
(but how many people do that, and rely specifically on quotes?)
- may break existing documents that incorrectly use \(oq to
specifically get the ASCII 0x60 ` output glyph
- typos of \(oq vs. \(cq are no longer obvious in ASCII output
(but ASCII is weak for detecting typos in the first place)
