The UTC has modified the conformance clauses and the text they reference. These changes clarify precisely what is tailorable in conformant implementations and what is not. The non-tailorable results are limited to interactions among a small set of well-defined core characters, such as CR, LF, NBSP, SP, and so on, where the semantics of the characters is bound up in how they linebreak.
The rules of the algorithm have been reordered and restated to allow free tailorability of most rules and the character properties they depend upon, except for those rules involving the core set.
Reviewers should note that all the rules were renumbered, with a correspondence table between rule numbers in the Modifications section.
The default assignments to linebreak classes for certain South East Asian scripts have been adjusted to make it easier to interface with the kind of external algorithm needed to handle these scripts correctly.
A few rules have been added or modified to better reflect the behavior of currency characters (and similar characters) in numbers, and the use of words like "person(s)".
Other minor adjustments have been made to line break classes, rules and text.