John Cowan wrote:

> Thomas Chan scripsit:
> > That doesn't hold up for multi-line text. e.g., at an archive of
> > newspaper clippings[1], there is a clipping from a 1999
> issue (date not
> > indicated) of an article[2], where the four-line photo
> caption is a single
> > sentence paragraph, with a paragraph indent, text running rtl, and
> > embedded English text running ltr (but the list commas and
> open/close
> > quotes are still ltr style).
>
> Okay, I'm sold. Now: why is it being done this way?

Whatever the reason, it's nice to see that Unicode got it right, for once.
:-)

UnicodeData.txt:
...
3001;IDEOGRAPHIC COMMA;Po;0;ON;;;;;N;;;;;
...
300C;LEFT CORNER BRACKET;Ps;0;ON;;;;;Y;OPENING CORNER
BRACKET;;;;
300D;RIGHT CORNER BRACKET;Pe;0;ON;;;;;Y;CLOSING CORNER
BRACKET;;;;
...
(List comma is not mirrored in RTL text; Chinese quotes are.)

BidiMirroring.txt
...
300C; 300D # [BEST FIT] LEFT CORNER BRACKET
300D; 300C # [BEST FIT] RIGHT CORNER BRACKET
...
(The default glyphs for open/close quotes are swapped in RTL.)

So, all this would work out-of-the-box in a properly implemented Unicode
Bidirectional Algorithm.

Ciao.
Marco