The Unicode Consortium Discussion Forum

Date/Time: Sun Sep 4 04:55:13 CDT 2011
Name: Philippe Verdy
Report Type: Error Report
Opt Subject: PRI#205 (proposed addition of ELM)

The justification given for adding ELM is to allow overriding the bidi
class of a character encoded before it, when it has a strong or weak
LTR direction, and when it occurs after a character with strong (but
contextual) direction like European digits. The case is shown with "/"
that has Bidi class CS (weakly LTR) when it comes after a European
digit "0" to "9" with Bidi class "EN". The proposed solution is to use
ELM so that it changes the Bidi class CS into the Bidi class of the
previous EN character.

This effectively is the same as creating a new Bidi class, because it
requires all existing UBA implementation to be updated to accept the
new behavior.

I also agree with the comments given in L2/2011-306 (ELM comment) by
Matitiahu Allouche, following another comment in L2/2011-183... This
proposal in PRI#205 breaks the stability promissed.

None of the proposed 3 solutions are in fact acceptable if stability
is required. But a 4th solution is possible:

Why not instead encoding another '/' with an strong RTL bidi class
(the same as other Arabic digits), named for example "ARABIC NUMERIC
SLASH SEPARATOR".

Those characters may not be the subject of the same mirroring
behavior, if it ever applies to numeric dates inserted in an RTL
paragraph.

No change would be needed in the UBA, stability would be preserved and
we would get the expected rendered ordering.

It is infeasible, because it would require cloning an indefinitely large number of characters that would be visually identical to others; they would invariably be mixed up in text, resulting in unpredictable rendering.