i18n API collation question

# Nebojša Ćirić (13 years ago)

my implementation fails this collation test:

intl402/ch10/10.3/10.3.2_CE.js

for this pair (a+umlaut+underdot):

"ä\u0323", "a\u0323\u0308"

If I turn normalization on then test passes.

Mandatory normalization introduces higher processing cost (up to 30% slower in ICU). ICU team decided to avoid normalization for some locales where they don't expect problematic characters to occur.

My question is, do we want normalize all strings by default or not, in compare() method? I think we said no to default normalization at one of the i18n meetings, but I am not 100% sure.