Created attachment 181652 [details]
inseparable.html. Reproduced content for justification.
In the attached inseparable.html, U+3033 should not be separated from U+3035, but it is separated.
This bug is only reproduced on Mac, because other platforms don't expand between ideographs.
Requirements for Japanese Text Layout say not to separate the characters.
> Combinations of character classes which allow spaces to be inserted for line alignment, are described as a complete table in Appendix E Opportunities for Inter-character Space Expansion during Line Adjustment, following 3.9 About Character Classes.
In 3.9 About Character Classes, U+3033 and U+3035 are Inseparable characters (cl-08).
In 4th note in Appendix E.2 Notes:
> A third order opportunity exists for inter-character space expansion, to take up to a maximum of a quarter em space, with respect to the corresponding character size, between two consecutive inseparable characters (cl-08) which are of different kinds.
Therefore, we should not separate separate U+3033 from U+3035.
Line breaking also is occurred between U+3033 and U+3035.
Please watch inseparable-line-break.html.
Requirements for Japanese Text Layout say not to break line between the characters.
In 5th note in C.2 Notes:
> There is no line break opportunity between following couple of consecutive inseparable characters (cl-08) as follows:
> VERTICAL KANA REPEAT MARK UPPER HALF "〳", VERTICAL KANA REPEAT MARK LOWER HALF "〵"
> VERTICAL KANA REPEAT WITH VOICED SOUND MARK UPPER HALF "〴", VERTICAL KANA REPEAT MARK LOWER HALF "〵"
Created attachment 181653 [details]
inseparable-line-break.html. Reproduced content for line breaking.
Created attachment 181657 [details]
(1) line break opportunities need to be determined by ICU and not use a hardcoded escape around ICU such as Font::isUnbreakableCharactersPair;
(2) the JLREQ document  is not a W3C recommendation; it is a collection of input requirements being considered for preparing normative recommendations, such as CSS3 Text, the current draft of which defines the recommended behavior in ;
(3) the current Unicode Line Break class database marks U+3033 and U+3035 as ID (Ideograph) class, and not IN (Inseperable); in general, ICU and CSS3 Text make normative reference to this database for determining line break classes;
(4) there is already a pending patch in process  which will be adding line-break property support according to , so any change for JLREQ related line breaking should be handled as part of ;
mark as dependent on bug 89235 to resolve line break semantics for japanese
Thank you, Glenn.
Your advice is very helpful to me.
I will ask CSS guys and Unicode guys to follow JLREQ behavior.
Therefore, I currently remove review flag.
Unicode 6.3 will fix line break property for U+3035 to CM. It will be propagated when ICU incorporates new data from CLDR.
Please be prepared, ANY * CM will not break, and not to justify between them.