5 Proofs
The old shop proved its work twice, and never by asking the man who set it. First came the galley proof, a rough pull marked against copy in the reading closet. Then, after imposition and makeready, the pressman pulled a sheet at the press itself: the actual sheet, on the job’s own paper and ink, from the press that would run it. The master passed that sheet before he let the run begin.
He studied the sheet, not the manuscript, and not the compositor’s account of his own care. Intention leaves no mark on paper; only the press does.
Our shop keeps the custom. The devil sets the pages now. He is quick and tireless, and when he finishes he reports the work done. Sometimes it is. The evidence is the file on the disk, and the file does not care what anyone intended. So after every change we pull a press proof, and the proof puts questions to the file that the devil cannot answer for it. When the product is software, the press proof is the artifact a customer receives: the release installed fresh and run, the pages loaded as a stranger would load them, not the log that says the build completed.
5.1 What the proof asks
We still pull galley proofs, in their way: the prose gets read, the style gets checked line by line, the claims get held against their sources. But a galley proof only proves the words. The press proof proves the object, and it runs after every change, no matter how small the change looked. Small edits have wrecked whole runs.
Count the pages. The book has a length, and the proof knows it within a narrow band. A run forty pages short has dropped a chapter; a run forty pages long has set one twice. The count catches either before a reader can.
Measure the trim. The leaf has one size, and every page must hold it. A stray setting can shrink the block or swell the margins without a word of complaint, so the proof measures the first page, the last, and a handful between.
Confirm the faces are embedded. A book that borrows its type from whatever machine opens it will look different on every machine. The proof looks inside the file and confirms that every face travels with it.
Find the sentinel lines. The shop keeps a few ordinary sentences from the prose, sentences that survive only a real run: one from the opening chapter, one from the last, one from deep in the middle. If a sentinel is missing, the press set something other than the book, however handsome the result. Take the sentinels from the prose itself, never from the front matter; a title page survives almost any disaster.
Render every page and look at it, every one, not a sample. Each must carry ink, because a blank leaf where prose belongs still counts as a page, and no count will catch it. Nothing may touch the edge, because the knife takes half of any line that kisses the trim.
Follow every link in the plate list. Each entry claims a page; the proof turns to that page and confirms an image truly stands there.
None of this is subtle work, just counting, measuring, and looking, the oldest work in the shop. The proof never asks whether the book is good, only whether the file in front of you is the book.
5.2 Never proof yesterday’s sheets
The proof pulls a fresh sheet, every time. A proof examines whatever sheet you hand it, and it will bless a stale one as happily as a fresh one. Change a line, then proof last night’s run, and the proof comes back clean; you have checked yesterday’s work and learned nothing about today’s. The sheet it read predates the change.
So in this shop the pull and the proof are one motion. The press runs first, then the questions are put to what it just produced. Do not trust anyone, yourself least of all, to remember the order at the end of a long day. Build the rebuild into the proof itself, so that skipping it takes more effort than doing it. A proof that can be pointed at old sheets is not a proof. It is a ceremony.
5.3 The run that lied
One book taught this shop to render every page. A plate came in a fraction of an inch taller than the text block. The press took it without a warning, reported success, and shipped blank leaves where the rest of the book belonged, run after run, from that plate onward. Nothing in the machinery ever objected.
A person caught it, by looking: turning the rendered pages one at a time until white space appeared where prose should have been, then paging on through leaf after leaf of nothing. The fix took minutes; crop the plate a hair shorter than the block and the book came back whole. The habit of turning every page went into the proof that same day.
The press was not unusually faulty. Every press has failures nobody has met yet. A success report measures the press’s satisfaction with itself, and this press was entirely satisfied with a blank page. You have to look.
5.4 The copy on the shelf
One distinction outlasts publication day. The copy on the bench, the one the press just produced and the proofs just blessed, is not the copy a stranger will be handed. Between the run and the shelf come the gathering, the wrapping, the carting, the shelving, and any of them can hand the customer something other than what the press produced. None will mention it, because none knows. Proofing the bench copy proves the press. Proving the copy on the shelf belongs to release day, where this manual takes it up.