Just noticed that Integrity Test on a RM9 database when using Mac Ventura passes but when I opened the same database in Windows the Integrity Test fails with numerous errors. Running re-index corrects the problem in Windows but back on the Mac I have to re-index again. Has anyone else seen this?
I can’t add anything meaningful to the really good discussion of the problem referenced above by BobC. I do hope that the RM developers take this problem seriously. I could be wrong, but I suspect that it’s not going to be an easy problem to solve.
Except the problem might (or might not?) be easily solvable if RM abandoned its special collating sequence and instead used the standard SQLite collating sequence. RM’s special collating sequence sorts non-English letters as if they were the English letter they most resemble. For example, the Norwegian Å sorts as if were an English A, the German Ü sorts as if it were an English U, and the Spanish Ñ sorts as if it were an English N. This wreaks havoc for speakers of those languages who are using RM and who might wish the letters to be sorted according to their own native alphabets.
I don’t know if simply using the standard SQLite collating sequence would solve the sorting problem for other languages. I suspect it would not and that more heroic measures might be required to support proper sorting that is language specific. But I do wonder if using the standard SQLite collating sequence might solve the Windows vs. Mac collating problem.
I have written about RM’s collating problem before. Even if the use of the standard SQLite collating sequence would not solve all the problems for sorting languages other than English, I do think it would be a step in the right direction by separating out letters that really aren’t the same instead of treating letters that are really not the same as if they were.