CLDR doesn't modify anything but primaries in the root ordering. Particular languages may modify any of the levels, but I don't think anything is typically done except for primary and secondary (with the exception of Japanese, which is quite complicated).
Mark *— Il meglio è l’inimico del bene —* * * * [https://plus.google.com/114199149796022210033] * On Fri, Jan 27, 2012 at 13:51, Ken Whistler <k...@sybase.com> wrote: > On 1/27/2012 1:16 PM, Matt Ma wrote: > >> Hi, >> >> There are a few characters having no decomposition type defined in >> UnicodeData.txt, but they were assigned tertiary weight in >> allkeys.text as if the characters had decomposition type. Here are a >> few examples (version 6.0.0), >> >> ... >> > > U+A733, U+A732, U+1F1E6 were given tertiary weight as they were >> <compat>, while U+31B4 as it were<final>. >> > > Yep, that is all done deliberately, to make the default sorting a bit more > consistent. > The normative decompositions in UnicodeData.txt are only the starting point > for attempting to give more consistent default weights for collation. > > > >> Is this something documented outside of UCA? >> > > No, because it is only relevant *to* UCA. At least as far as documentation > written by the UTC is concerned. > > Well, I suppose it is also relevant to CLDR, because CLDR bases its > collation > tables on a tailoring of allkeys.txt from UCA. I don't know what > documentation > there may or may not be about the default treatment for tertiary weights > in CLDR. Somebody involved in the details of CLDR collation will have > to answer that one. > > --Ken > > > >