Where everything is tonight

The planets come from JPL’s Keplerian elements for approximate positions of the major planets, which is six elements and six rates per planet, fitted over 1800 to 2050. JPL says plainly that they are “not valid outside the given time-interval”, so the clock refuses to leave it — it stops at both ends of the scrub rather than carrying on into a century nobody fitted. Inside it, the published error runs from 15 arcseconds of heliocentric longitude for Mercury to 600 for Saturn, which at the scale of this figure is under a pixel for every planet on it.

The 24,045 asteroids and comets behind them come from JPL’s Small-Body Database, which publishes each object’s elements osculating at one epoch — a snapshot of the orbit it is on today rather than a fit across two centuries. Scrub far from that epoch and the belt is telling you the shape of the solar system rather than the position of any one rock in it.

Drag the figure, or press the arrow keys, or take one of the three presets. Pole is the view every diagram of the solar system is drawn in, straight down the ecliptic’s north pole. Edge is the one that shows what that view is hiding: seen from within the plane, the eight planets collapse onto a line about seven degrees thick, and everything that does not is the point — Pluto seventeen degrees out of it, Halley going round the other way at a hundred and sixty-two, and the belt revealed as a torus rather than a ring.

The positions were always three-dimensional — every body here publishes a longitude of the ascending node, which is more than can be said for the exoplanet system figures on this site, drawn perpendicular to their orbital plane because their catalogue publishes that angle for not one planet in it. What this page used to do was compute all three coordinates and then throw the third one away at the last step. It no longer does.

The projection is orthographic rather than perspective, which is a decision and not a shortcut: nobody has ever stood anywhere that the solar system looks like a perspective drawing from, and a perspective camera would make the scale bar a lie. Here a horizontal distance on the screen is exact at every angle the figure can be turned to. What tilting costs is the vertical, foreshortened by the sine of the tilt, exactly as it should be.

The radial scale is linear, as it has to be: a logarithmic radius turns an ellipse into a shape that is not one, and the shape is what tells you that Halley and Venus are not the same kind of object. So the way in is the field control rather than a distorted axis, and each step in drops the outermost orbit.

The named bodies and their orbits are SVG, rendered on the server: they are in the HTML, they are there with JavaScript switched off, and a crawler reads them. The 24,045 small bodies behind them are not, and cannot be — each one has to be moved every time the clock ticks or the view turns, and moving one means solving Kepler’s equation for it. That is Newton’s method, three or four iterations, on twenty-four thousand entirely independent bodies, sixty times a second.

Where the browser has WebGPU, that happens on the graphics card: the elements are uploaded once and every frame after that costs a sixty-four-byte write for the clock and the camera. Where it does not — roughly a fifth of browsers — the same arithmetic runs in JavaScript and the belt is drawn on a 2D canvas instead. The figure says which one it used. Nothing else on the page depends on the answer.

Mercury, Venus, Earth, Mars, Jupiter, Saturn, Uranus, Neptune — and, past them, the named small bodies the figure draws: Apophis, Bennu, Ryugu, Eros, Encke, Vesta, Ceres, Pallas, 67P, Halley, Pluto, Haumea, Quaoar, Makemake, Eris, Sedna. Each has a page with its own measurements, and every element this figure moves it by is on that page.