Galley Proof

Galley Proof

Text arrives carrying things nobody typed. A no-break space out of a word processor. A zero-width space picked up crossing a CMS. A Cyrillic о sitting inside a Latin word in an address someone sent you. A soft hyphen that stays invisible until the line breaks and then prints a hyphen in the middle of a word. None of it shows on screen, all of it survives the paste, and you find out later — in a diff, in a search that returns nothing, in a name that will not match the one in the database.

Every editor can show you invisible characters as a wall of boxes. That is a lint, not a proof. This is the other tradition: the galley proof a printer pulled off the tray before a page existed — a single column of type, a wide margin, and corrections written twice. A slash through the offending sort where it stands, and out in the margin what to do about it, with a stroke after it.

Paste anything into the Manuscript and the sheet sets itself.

On the sheet	Reading
Red slash through a character	This is to be changed or taken out
Red caret	A character with no width of its own is standing here
Blue dots under a character	Stet — it was marked, and it stays
Margin: dele loop	Take it out
Margin: #/, -/, '/	Substitute this, and stop reading at the stroke
Margin: close-up arcs	Close the space up
Margin: ?	A query to the author — the tool cannot fix this one
Margin: stet	Left standing
Line figure in red	This line carries at least one mark

Click any mark — in the type or in the margin — to swap it between set and stet. The schedule of marks under the sheet does the same thing wholesale: one line per kind, with its count, its codepoint, and set / stet for all of them at once. Hovering a line of the schedule dims the sheet down to just that kind, so you can see at a glance where the no-break spaces actually are. Then Revise shows the corrected copy, and Copy the revise puts it on the clipboard.

Nothing is uploaded. The analysis, the marks and the revise all happen in the page, which is the only reason it is safe to paste a real address into it.

What gets marked, and what is left alone

The defaults are the judgements a working copy editor would make, not a lint’s idea of purity.

Marked and set	Marked but left standing
Zero-width characters, BOMs, Hangul fillers	Curly quotes, apostrophes, primes
Bidi controls — the trick behind reversed filenames	En and em dashes, minus signs
Soft hyphens, control characters	The ellipsis character
No-break, thin, hair and ideographic spaces	Fullwidth Japanese punctuation
Double spaces, trailing spaces, carriage returns	Tabs — a pasted table has to survive
Fullwidth letters and figures, halfwidth katakana	Variation selectors
Look-alike Cyrillic and Greek letters in Latin words	An ideographic space indenting a Japanese line
Accents stored apart from their letter	The replacement character — only the author can restore it

An ideographic space opening a line is a Japanese paragraph indent and stands; the same character in the middle of a Latin sentence is junk and is set. Fullwidth １２３ is almost always an accident; fullwidth （） almost never is. A look-alike letter is only a look-alike inside a word that is otherwise Latin — an ordinary Russian sentence is left in peace.

The watermark

Under the counts there is a line that reads the sheet the way a conservator reads a watermark: hold the paper up to the light and it tells you which mill it came from. No-break spaces together with curly quotes and soft hyphens mean a word processor. Zero-width characters mean the copy crossed a web page. Fullwidth forms mean a Japanese input. Carriage returns mean a Windows editor. A replacement character means an encoding the text did not survive. It is the same evidence that produced the marks, read for provenance instead of correction.

Key	Action
J / K	Move to the next or previous mark
S / F	Stet it · set it
Return	Toggle the selected mark
C	Copy the revise

You can also drop a .txt file onto the sheet, or paste straight onto the galley without going back to the Manuscript.