L2/04-291 Date/Time: Thu Jul 15 04:01:37 CDT 2004 Contact: kentk@cs.chalmers.se Report Type: Error Report Subject: mirroring, problem 1: arrows Many of the math symbols are set to be mirrored in RTL contexts. However, none of the arrows are set to be mirrored. Since arrows are (most often) used as (relational) math symbols too, this seems inconsistent. Math formulas would be written LTR in most cases. But explanation of symbols may easily occur in RTL scripts, with a quote of the symbol. In such cases one would expect all relational(?, or all dyadic; what about not sign?) non-symmetric math symbols to consistently be mirrored or consistently not be mirrored, whether arrow or other. -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- Date/Time: Thu Jul 15 04:04:24 CDT 2004 Contact: kentk@cs.chalmers.se Report Type: Error Report Subject: mirroring, problem 2: decompositions Negated math symbols have a decomposition, using 0338. Consider U+2260, NOT EQUAL TO, which is listed as mirrored with no mirror best fit (not even itself, oddly). It has a decomposition into <003D, 0338>. (The variant with a vertical stroke is not a problem here.) The decomposition has no mirroring, and is perfect as it is. Is it really expected that the stroke slant should be mirrored in RTL contexts? If so, 0338, at least, should have the "mirrored in RTL contexts" property. If not, negated math symbols should be listed as having a perfect (not just "best fit",or even missing) mirror character. The latter then goes for several "partially negated" math symbols that have no decomposition (many of them would have a "perfect mirror" char), as well as negated arrows (see mirroring problem 1). -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- Date/Time: Thu Jul 15 04:28:18 CDT 2004 Contact: kentk@cs.chalmers.se Report Type: Error Report Subject: mirroring, problem 3: BidiMirroring.txt data file format and extenstion of mirroring data supplied The characters that have no mirror character should be listed not just as comments but as data lines. E.g.: 22AA; # TRIPLE VERTICAL BAR RIGHT TURNSTILE The characters that have a "best fit" mirror character (the few that may be left after dealing with problem 2) should be listed as being approximate in the data part (not [just] in the comment part). E.g.: FF62; FF63 # HALFWIDTH LEFT CORNER BRACKET Some mirrored characters may occur in italics or oblique. (Most do not, like the math symbols whose glyphs are never inclined even in italics/oblique styles.) E.g. parentheses. It would be a good idea to note that for italics/oblique, actual glyph mirroring (along a vertical axis) for these characters is imperfect, but substituting with the mirror character('s glyph) is best practice. E.g.: 0028; 0029 # LEFT PARENTHESIS (use some other name for the tag if you like). Addendum to mirroring problem 2: some characters that currently have the mirrored property may now have themselves as perfect mirrors; they need not keep the mirrored property. -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --