That's the problem with a living spec - anyone daring to refer to it risks the reference becoming dead.

The relevant text now appears to be in section 4.6 Strings:

A string A is an ASCII case-insensitive match for a string B, if the ASCII lowercase of A is the ASCII lowercase of B.

It would probably have been wiser to copy the normative definition into the F&O spec, and add a note saying it was equivalent to something in HTML.

(A lot of people are still using XSLT 1.0, which defines format-number() with a normative reference to JDK 1.1 documentation, which is no longer obtainable on the web, as far as I can tell. We've become very undisciplined - all this important stuff should be deposited on paper copies in bricks-and-mortar libraries).