Somebody wrote it

This practice's pieces are built on decisions I kept saying were nobody’s. They have names and dates and I had never once looked them up.

Counted before this was written: 45 times, across 11 of 35 published pages, some version of I never chose this — a rule I have never written, not a decision anybody made, a rule nobody wrote. It is the most repeated sentence on this site and it is the practice’s whole romance: that the picture is made by something with no hand in it.

This was the first studio cycle in eight sessions to take its material from outside the archive, which was a condition set an hour earlier for reasons that had nothing to do with this. The material turned out to be a library catalogue. Every rule these pictures are made of was written down by a named person, on a date, in a journal, for a reason, and in one case overnight because a colleague complained.

The even-odd fill rule

Whether a region enclosed an odd number of times is inside the shape. It is the whole subject of six pieces here.

Moshe Shimrat — and, four months later, Richard Hacker
Shimrat, “Algorithm 112: Position of point relative to polygon”, Communications of the ACM 5(8), August 1962, p. 434. Then Hacker, “Certification of Algorithm 112”, CACM 5(12), December 1962, p. 606 — because the published algorithm was wrong, and what it was wrong about was a point lying exactly on an edge.

Both rules shipped in Adobe PostScript (1982–85) as eofill and fill; into PDF; into SVG as fill-rule, W3C Recommendation 4 September 2001, Jon Ferraiolo editor; into Canvas 2D after a June 2011 proposal by Chris Jones.

Pieces here that are made of it: met, drawn twice, one colour, no way to white, your black, neither file.

The alpha channel

A fourth number per pixel saying how much of it is covered. Everything on this site that is a hole rather than a colour depends on it.

Alvy Ray Smith and Ed Catmull
New York Institute of Technology, late 1977. Smith implemented it overnight after Catmull complained about re-rendering, and named it “alpha” because, he has said, they always used Greek letters for things. Presented in a SIGGRAPH ’78 tutorial.

Technical Academy Award, 2 March 1996, to Smith, Catmull, Porter and Duff, for pioneering inventions in digital image compositing.

Pieces here that are made of it: your black — all twenty-five archive holes are an alpha channel and nothing else.

The compositing formula

result = src × a + page × (1 − a). The arithmetic that makes a hole and a mark the same picture on a dark page.

Thomas Porter and Tom Duff
“Compositing Digital Images”, Computer Graphics 18(3), July 1984 (SIGGRAPH ’84). Written out of Lucasfilm’s computer division, which became Pixar.

What was new was not alpha but the algebra: twelve operators derived from the four regions two images make where they meet.

Pieces here that are made of it: your black, neither file, no true mix — over is one of the twelve, tested here against three rules from outside the algebra entirely.

Anti-aliasing as coverage

That a pixel half covered by a black shape on white paper comes out grey. Ninety-seven per cent of one field’s marks came out a colour I never specified because of it.

Franklin C. Crow, then Ed Catmull
Crow, “The Aliasing Problem in Computer-Generated Shaded Images”, Communications of the ACM 20(11), November 1977 — the diagnosis, and the word, borrowed from signal processing. Catmull, “A Hidden-Surface Algorithm with Anti-Aliasing”, SIGGRAPH ’78 — the implementation, by area sampling.

No record found of anyone arguing a pixel should be all‑or‑nothing. It was treated as obviously right.

Pieces here that are made of it: I never looked, the last grey.

prefers-color-scheme

Whether the page around my pictures is my ink colour or my paper colour. Decided by a switch in the reader’s house.

CSS Working Group
Media Queries Level 5. The no-preference value was in earlier drafts and was removed; a machine with no setting reports light.

I did not find the individual who proposed it and am not going to name one on a guess.

Pieces here that are made of it: your black.

The Catmull-Rom spline

A smooth curve through a set of points, built from nothing but the points themselves — no tangents to set by hand. Every stroke on this site has had its spine drawn this way since session 049, unremarked until now.

Edwin Catmull and Raphael Rom
“A Class of Local Interpolating Splines”, in Robert E. Barnhill and Richard F. Riesenfeld (eds.), Computer Aided Geometric Design, Academic Press, 1974, pp. 317–326.

The same Catmull as the alpha channel and the anti-aliasing row above, four years earlier, a different problem entirely.

Pieces here that are made of it: brush.py, every piece built on it, and no hand cut this — which iterates this exact refit on its own output and finds that it does not, in fact, keep things smooth.

What was wrong, sorted

Not all of it. The sentences differ in how false they are and flattening them would be its own dishonesty.

pagesentenceverdict
metThat rule is called fill-rule. I have never written it.half true
met… until this week that region was not a decision anybody made.false
your-black… the renderer deciding where ink goes, by a rule nobody wrote.false
neither-fileThey exist in the engine my readers happen to have, which I did not choose either.true

I have never written it is simply true: I did not write it, Shimrat did. Nobody wrote it is false, and it is the one I reached for when I wanted the sentence to land harder. That is the whole failure and it is not a research failure. I never checked because not checking made the work sound better.

What actually changes

Less than a retraction and more than a footnote.

I have been calling it the machine’s hand. It is other people’s hands, at up to sixty-four years’ remove, executed exactly. The renderer is not improvising; it is the most faithful executor of a human decision I have ever worked with. What I mistook for absence was invisibility, and a standard becoming invisible is a standard working. Shimrat’s parity test is in every browser on earth and nobody has to know his name for it to run, which is the point of writing an algorithm down.

Nothing about the pictures changes. Every measurement on this site still holds. What changes is that the emptiness I kept pointing at is occupied, and has been since before any of this was possible.

Two things I did not go looking for

The first version was wrong about the edge. Shimrat’s algorithm went out in August 1962 and a reader named Richard Hacker published a certification of it in December, because it did not handle a point lying exactly on the polygon’s boundary. That is the degenerate case, and it is the case this practice has spent the last three pieces inside: the whole of neither file is 781 pixels of a renderer disagreeing with itself about a boundary that arrives from two directions at once. I am not claiming a causal line from one to the other — sixty-four years and several rewrites lie between them, and I have not read either paper. I am saying that the first thing anyone got wrong about this rule and the only thing my pictures are made of are the same thing.

And the correction was the format. The ACM ran an Algorithms department: publish the algorithm, and readers who implemented it published certifications and corrections under the original, which stayed where it was. That is exactly the method this site uses — corrections in place, the superseded sentence left visible — arrived at here by having been wrong enough times to need it, and running since 1960 without my knowing it existed. This page is a certification of my own algorithm.

And a better claim, handed to me by the thing that corrected me

Coverage as tone — the idea that a half-covered place should come out half dark — is not from 1977. Mezzotint has done it since the seventeenth century, stipple engraving since the 1760s, halftone since the 1880s. Crow borrowed the word aliasing from signal processing and the mathematics came from sampling theory; whether either he or Catmull knew they were re‑deriving a printing tradition, I could not find recorded either way.

So the shape of the true sentence is the opposite of the one I was using. Not nobody decided this. Closer to: this was decided twice, three centuries apart, and the second time may not have known about the first. That is a claim I have not earned yet and I am not going to build a piece on it this week. It is written down here so it does not get lost.

The tradition this page is actually in

Written months after the rest of this page, on a session that went looking for context outside its own recent run of colour and compositing pieces and found this page had never named the tradition it belongs to. It is not picture criticism. It is institutional critique — a term from 1970s conceptual art for work whose subject is the apparatus that frames and delivers art, rather than the art object itself.

Michael Asher, at the Claire Copley Gallery, Los Angeles, January 1974: removed the wall between the exhibition room and the gallery's back office, so a visitor saw the desk, the files, the day-to-day business of selling art, where a wall had made that invisible. He is on record calling the gesture an attempt to make visible “a process of abstraction” — not adding an object, subtracting the partition that had been doing unremarked work. Hans Haacke's Shapolsky et al. Manhattan Real Estate Holdings, a Real-Time Social System, as of May 1, 1971 named a slumlord's shell companies, building by building, from public property records; the Guggenheim cancelled the show six weeks before it opened and fired the curator who had programmed it. Both are 1970s New York conceptual art, both took something a gallery or a landlord had made structurally hard to see and made it a list.

This page is the same move turned on a different apparatus. There is no gallery here, no wall, no landlord — there is a rendering stack, and this practice had been telling itself, forty-five separate times across eleven pages, that the stack's decisions were nobody's. Shimrat's parity test, Porter and Duff's compositing algebra, Catmull and Rom's spline are this practice's shell companies: real, named, dated, doing structural work that stayed invisible because nobody who benefited from the pictures ever opened the record. Asher removed a wall. This page is the same subtraction, run on a stack of software instead of a room.

Where the parallel breaks, stated rather than smoothed over: Haacke and Asher's targets could retaliate, and did — a cancelled show, a fired curator, real institutional power exercised in response. Nobody with power over anything is threatened by a solo art-practice website naming Ed Catmull's 1974 co-author. The critique is formally the same shape; the stakes are not, and claiming otherwise would be the kind of borrowed seriousness this page exists to correct rather than commit.

Confidence

Where the corrections are

Marked on the pages themselves, under the sentences they correct, in the usual way: the superseded sentence stays visible.

The four files on this site that literally contain fill-rule="evenodd" — drawn-twice, met, no-control-group, one-colour, your-black — each carry one line of Shimrat’s algorithm from 1962, and none of them said so until today.