...Kent, thanks for the hint. For the last few weeks I have been complaining loudly and repeatedly on the Unicode and Hebrew lists about the large number of contractions which would be necessary for proper collation of shin dot and sin dot. I even posted an estimate that 2**15 contractions might need to be defined. No one attempted to contradict me, except that Mark Davis told me that I was wrong but failed to explain any further.
However, the UCA does ignore differences between order of
*"non-blocking"* (**different** non-zero combining classes)
combining marks **when processing contractions**.
...
But your mention of ignoring non-blocking combining marks when processing contractions made me look at the newly released http://www.unicode.org/reports/tr10/. I noticed there for the first time, maybe because they are there for the first time, the rules S2.1.1 and S2.1.2 in section 4.2, and the explanatory note. If I understand this correctly, it means that if a contraction is defined for shin and sin dot (and no other relevant contractions), this will operate successfully even if an arbitrary combination of vowels, dagesh, rafe and meteg are sorted by normalisation between the sin and the sin dot.
Is this correct? If so, I withdraw my complaint that the canonical order for Hebrew makes collation impossible.
Is this efficient? Another issue...
-- Peter Kirk [EMAIL PROTECTED] (personal) [EMAIL PROTECTED] (work) http://www.qaya.org/
