Links and marks hung on text the engine has already placed: a band behind a line, an underline, a boxed cell — every rectangle taken from what the layout and the table report afterwards, none of it measured by hand.
The hard part of a link on text is not the link: it is knowing where the line ended up. So the text and the table are drawn first and then asked where each line and each cell landed, and the marks are hung on the rectangles they report.
A pale band is painted behind a paragraph's first line, one line is underlined and made clickable, and one table cell is boxed and made clickable. Nothing on the page is measured by hand, which means the marks stay right when the text is edited, translated or set in another font.
The same document is written twice, once in each language. Both files are shown in full, with the code coloured on the server, and can be copied in one click.
The same document is written in each of these languages. Open the one you want to look at.
| Before | 110,681 bytes |
|---|---|
| Now | 40,181 bytes |
| Saved | 63.7 % |
Not one mark on the page changed. See every document measured.
Thirteen slides, each stating the effect it arrives by — all twelve the standard defines — how long that effect takes and how long the slide stands before the next one comes. The file opens full screen, which is the only place a reader reads any of it.
Four pages that state what happens the moment the file is opened: which page it opens on and how much of the window that page fills, the bookmarks shown beside it, the pages two at a time, a print dialogue already filled in, and the address its relative links are read against.
Pages that link out of themselves and to each other.
Pages numbered by their labels rather than their position.
A page of PDF417 codes, drawn as stacked bars.
A page of Aztec codes, drawn as squares.