UAX #14 rev. 57, Section 9.1: the rewritten rules do not give the same result as LB9 and LB10

Section 9.1 says the rules from LB11 on can be rewritten so that “the overall result of the algorithm is unchanged”: replace LB10 by “× CM” and add a CM copy of every rule with AL on its left side. Applied literally, the rewrite differs from Section 6.1 in three cases: a space before a combining mark (no break instead of the LB18 break), ZWJ, which LB9 treats like CM but Section 9.1 does not mention, and AL-to-CM copies that also match marks attached to other bases. On 1,099 of the 19,346 lines of LineBreakTest.txt, 58 lines give a different result.

Consequence

U+0061 LATIN SMALL LETTER A, U+0020 SPACE, U+0301 COMBINING ACUTE ACCENT, U+0062 LATIN SMALL LETTER B: Section 6.1 gives a break between the space and the accent, the rewrite of Section 9.1 gives none.

Proposed correction

Keep the treatment of SPACE, ZWJ and combining marks that already have a base in Section 9.1, or qualify the sentence that the overall result is unchanged.

Issued by

Unicode Consortium — Unicode Technical Committee

Act

UAX #14 revision 57 (Unicode 18.0.0)

Place

Section 9.1

Record

Discrepancy

Class

Text of the act

Where it arose

The text of the act

Arxo's conclusion

Confirmed

Official text

Unchanged at last check

Last checked

2026-10-31

Evidence

Reproduced by execution. Quotations are verbatim from the official material; translations are ours and marked.

However, it is possible to make these rules unnecessary by rewriting all the rules from LB11 on down so that the overall result of the algorithm is unchanged.

Replace LB10 by the following rule: × CM

For each rule containing AL on its left side, add a rule that is identical except for the replacement of AL by CM, but taking care of correctly handling sets of alternate line break classes.

How to check it without Arxo's tools

1. Open UAX #14 revision 57 (Unicode 18.0.0), Unicode Consortium and find Section 9.1. Compare the text with the quotation above. A copy is kept in this register: 296df5332aefe695-tr14-57.html (SHA-256 296df5332aefe695b4951b511288b118bc6a7b43db7fadfd60f6062096acc7c0, copy of the official file).
2. The finding holds if the quoted passages read as shown and the comparison gives the result stated at the top of this page. If it does not, write to us with the record ID.

How this finding was made

AI agents read the annexes and the data files and compared them; the examples were run through published implementations and through rule models of both readings. An AI coordinator checked each place against the full annex and the trackers, and two further AI models checked the report before filing.

Responsible institutions

• Unicode Consortium — publishes the annex and the Unicode Character Database

Reported to

Responses

No response recorded.

History

1. 2026-10-31Confirmed by Arxo
2. 2026-10-31Entered in the register
3. 2026-10-31Reported to Unicode Consortium (web form)

Cite this record

Arxo Law, Arxo Findings Register, ARXO-2026-0455, https://arxo.io/findings/ARXO-2026-0455/ (status as of 2026-10-31). Data: ARXO-2026-0455.json.