Bidi and Empty Parentheses

[Richard Wordingham]

On Tue, 1 Mar 2022 13:06:28 -0700
Doug Ewell via Unicode <unicode at corp.unicode.org> wrote:

> Andreas Prilop wrote:
> 
> > You should not use the embedding characters any longer.
> > After 15 years of Bidirectional Algorithm, they finally discovered
> > that the embedding characters (as well as their HTML and CSS
> > equivalents) do not work as desired.  
> 
> This puzzles me. Character encoding is engineering, not natural
> science.
> 
> If some mechanism is defined incompletely or erroneously, the
> definition can be corrected. If some font or rendering engine doesn't
> handle a mechanism correctly, it can be updated.

This is not certain when there is accumulated text that relies on the
current behaviour.

> Very little of this falls into the category of "it worked for 15
> years, or so we thought, but we've discovered a case where it doesn't
> work, so now we have to abandon the whole thing."

The actual assertion appears to be that the LRE and RLE mechanisms
should be replaced by LRI and RLI, and in this case, it has been
proposed that the HTML for the former should now be interpreted as HTML
for the latter.  I struggle to see the equivalence, partly because UAX
#9 is still practically unintelligible.

In the example given by Eli, namely "RLE x ( ) y PDF", the point of the
RLE is to reliable force the paragraph direction to be right-to-left.
This is for a test case; the query was prompted by an example where the
only non-default higher level protocol applied to the definition of a
paragraph, and a previous line caused the line of interest to have
right-to-left direction.

> This reminds me of the discussion two weeks ago about Arabic
> presentation forms, in which it was explained (again) that font
> and/or rendering engine inadequacies were considered justification
> for using these non-preferred forms instead of real Arabic letters.

The 'discussion' was a defence of using them instead of attaching an
image.

Richard.