PostScript and the Raster Image Processor

Every other printer in this batch is told where the dots go. The computer decides, in that printer's own dot grid, using that printer's own escape codes - so a document is tied to the machine it was written for, and a better printer needs a new driver and gives you nothing you did not already have. John Warnock and Chuck Geschke had built Interpress at Xerox PARC and left in 1982 to sell the idea properly. PostScript, released in 1984, sends a PROGRAM instead of dots: moveto, curveto, fill, show, in coordinates that mean the same thing on a 300 dot-per-inch office laser and a 2540 dot-per-inch imagesetter. A raster image processor inside the printer executes it and produces the bitmap. That is why the Apple LaserWriter of 1985 had a 12 MHz 68000 and one and a half megabytes of memory - more computer than the Macintosh plugged into it. A 300 dpi A4 frame buffer is over a megabyte on its own and there is nowhere else to keep it. And the job is hard real time: once the paper is moving a laser engine cannot be paused, so a RIP that falls behind does not print slowly, it loses the page. That is the entire content of PRINT OVERRUN. Two things fall out of it that shaped everything after. Outlines scaled at print time are not enough on a coarse grid - at 300 dpi an eight point stem rounds up while a ten point stem rounds down, so the smaller type prints bolder - which is why fonts carry hints saying what must survive quantisation. And a page that is a program is a page nobody can predict: you cannot know how long it will take, or whether it terminates, without running it. PDF, in 1993, kept the whole imaging model and deleted the loops, and that is the document you are reading on a screen today.