Re: diagnostics: prefer "…" to "..." if the locale supports it

[Hans Åberg]

> On 12 Oct 2019, at 08:22, Paul Eggert <address@hidden> wrote:
> 
> On 10/11/19 10:35 PM, Akim Demaille wrote:
>>> For what it's worth, the Wikipedia manual of 
>>> style<https://en.wikipedia.org/wiki/Wikipedia:Manual_of_Style#Ellipses>  
>>> says, "Wikipedia's style for an ellipsis is three unspaced dots (...); do 
>>> not use the precomposed ellipsis character (…) or three dots separated by 
>>> spaces (. . .)."
>> I expect this to be for English.
> 
> Yes it is. In contrast, the French Wikipedia manual of style 
> <https://fr.wikipedia.org/wiki/Wikipédia:Conventions_de_style> uses ‘…’ 
> U+2026 HORIZONTAL ELLIPSIS without comment, which suggests that the 
> single-character ellipsis is more common in French.
> 
> An ellipsis in a proportional font looks fine, but in a monospace font it 
> looks odd in English – particularly in a context where an underscore would be 
> plausible too. Part of the problem is that ellipses look like (fuzzy) 
> underscores in some fonts. Part of the problem is that in some systems, 
> unknown characters are displayed as underscore, so that if the font lacks an 
> ellipsis U+2026 is displayed as an underscore which is doubly confusing. 
> (This may be the problem that Hans reported in this thread.)

It is another issue, unrelated to the underscore problem, namely that some 
fonts are so smart that they can render an ASCII ellipsis as the Unicode one, 
so they become visually indistinguishable. This happened with the subject title 
of this thread, so translating into ASCII, it looks like the non-sensical
  diagnostics: prefer "..." to "..." if the locale supports it
but in reality both looking the Unicode ellipsis “…".

> Since the confusion might be locale-dependent, perhaps all that’s needed is 
> to modify the translator comment to not insist on U+2026 so strongly. That 
> is, instead of saying “TRANSLATORS: use the appropriate character (e.g. "…") 
> if available.” it could say something like “TRANSLATORS: Use an ellipsis 
> appropriate for your language, remembering that "…" (U+2026 HORIZONTAL 
> ELLIPSIS) sometimes misdisplays and that "..." (three ASCII periods) is a 
> safer choice in some locales.”

I use Unicode characters in a program with a lot math characters, and at first 
I experimented with ASCII equivalent combination sequences alongside, but it 
became too complicated. This has also been noticed on the ConTeXt mailing list, 
which wants to be backwards compatible with TeX, but one arrives at 
ambiguities, such as “||” might mean two bars in some contexts but a double bar 
‖ U+2016 in others.

So I tend to think that if one calls for ASCII then one can use such 
combination sequences, but otherwise, if using Unicode, it one can just as well 
migrate using it. But that migration takes some work.