THE FOLD / CHEAT / NOCLIP / THE ZERO WIDTH JOINER
THE ZERO WIDTH JOINER
characters with no shape at all
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
Some characters have no shape at all. They occupy no width, print nothing, survive copy and paste, and change what a string is without changing anything you can see.
LIT verified live. five invisible code points — ZERO WIDTH SPACE, ZWNJ, ZWJ, WORD JOINER and ZWNBSP — inserted into the middle of a word give a string that is not equal to its visible twin in 5 of 5 cases, and all 5 belong to Unicode’s format category,
LIT verified live. five invisible code points — ZERO WIDTH SPACE, ZWNJ, ZWJ, WORD JOINER and ZWNBSP — inserted into the middle of a word give a string that is not equal to its visible twin in 5 of 5 cases, and all 5 belong to Unicode’s format category,
\p{Cf}. Trimming removes only 1 of them and leaves 4. The family emoji 👨👩👧 is 5 code points of which 2 are zero-width joiners; strip them and the same picture becomes 3 separate people.2 HOW IT WAS WEAVED · AI + HUMAN
Zero-width formatting characters are ordinary Unicode — ZWJ is how emoji sequences are built, ZWNJ is required to write Persian and Hindi correctly. They are not an exploit; they are typography.
AVAN (AI) tested the format category rather than eyeballing invisibility, because ‘invisible’ is a rendering claim and
AVAN (AI) tested the format category rather than eyeballing invisibility, because ‘invisible’ is a rendering claim and
\p{Cf} is a checkable one — all 5 match. The uncomfortable number is the trim result: only 1 of the 5 is stripped by the usual whitespace trim, because JavaScript counts just U+FEFF as whitespace. A defence that sanitises by trimming removes almost none of the class while looking like it did something.3 ONE DIMENSION
Five characters you cannot see. Five strings that are not equal.
4 TWO DIMENSIONS · INTERACTIVE
Insert an invisible character and compare the two words.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object.
AVAN’s addition (the inverse-companion): the forward reading is that invisible characters should be stripped. The inverse is that ‘invisible’ is not a property of a character, it is a property of a font’s decision not to draw it — and the same characters are load-bearing typography in other scripts. Strip them and you break Persian; keep them and two identical-looking usernames are different accounts. Read backwards, there is no sanitising rule that is correct for every writing system at once, which means the choice is a policy about whose text you are willing to handle.
LIT five invisible code points - ZERO WIDTH SPACE, ZWNJ, ZWJ, WORD JOINER and ZWNBSP - inserted into the middle of a word give a string that is not equal to its visible twin in 5 of 5 cases, and all 5 belong to Unicode's format category; trimming removes only 1 of them and leaves 4, while the family emoji is 5 code points of which 2 are zero-width joiners, so stripping them turns the same picture into 3 separate people
FIG Zero-width formatting characters are ordinary Unicode - ZWJ is how emoji sequences are built, ZWNJ is required to write Persian and Hindi correctly. They are not an exploit; they are typography. AVAN tested the format category rather than eyeballing invisibility, because 'invisible' is a rendering claim and the category is a checkable one. The uncomfortable number is the trim result: 4 of 5 are stripped and one is not, so a defence that sanitises by trimming leaves a hole, which is worse than removing none.
FIG Zero-width formatting characters are ordinary Unicode - ZWJ is how emoji sequences are built, ZWNJ is required to write Persian and Hindi correctly. They are not an exploit; they are typography. AVAN tested the format category rather than eyeballing invisibility, because 'invisible' is a rendering claim and the category is a checkable one. The uncomfortable number is the trim result: 4 of 5 are stripped and one is not, so a defence that sanitises by trimming leaves a hole, which is worse than removing none.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of NOCLIP · David Lee Wise (ROOT0), with AVAN