Date/Time: Thu Jul 28 02:35:06 CDT 2011
Contact: matial@il.ibm.com
Name: Matitiahu Allouche
Subject: Additional comments on proposal for ELM L2/11-183

1) The proposal should define if ELM is affected or not by LRO and RLO. I think it should not. This point could be moot depending on the answer to the next comment.

2) The proposal says that "the character behaves as if it were an LRM whenever the Bidi embedding level is L, and behaves as if it were an RLM whenever the Bidi embedding level is R." It is not clear to me whether the bidi embedding level is the paragraph level or the current embedding level (as modulated by occurrences of LRE, RLE, LRO, RLO and PDF). Both options have their own usefulness, but I think the first one is more useful, while the style of the proposal rather hints to the latter. In any case, the spec should say that the embedding level does not include resolution of implicit levels (what is shown in Table 5 of UAX#9).

3) In a clean design, ELM should consititute a bidi class of its own. I am aware of the Unicode Character Encoding Stability Policy stating that “The Bidi_Class property values will not be further subdivided." However, the proposed bypass of adding rule W0 raises all the compatibility problems that adding a bidi class would raise, and it is incredibly ugly in design. If we are looking for a formal loophole to create a new bidi class value without calling it a "subdivision", what we have here is a new class for a new character, so it is an addition rather than a subdivision. I strongly suggest to do the right thing (adding a bidi class value for ELM) rather than trying to cover the perpetration with appearances.