I fired up Chrome's dev tools to figure out how it works.
The JS that's doing the work is at http://imgs.xkcd.com/static/waldo.js. On page load, it sends a GET request to http://umwelt.xkcd.com/story/ghenkEggov8 with params w for the browser width in pixels, h for the height and r with the referrer. The rest of the qualifiers (geoloc and browser (user-agent)) can be determined on the server-side.
As a response, you get back JSON with the alt-text, image URL, and positioning info, which is then used to fetch and position the image.
The thread on xkcd forums has screenshots of a lot of the different comics. It's crazy how many variations there are. The one in hebrew (for those in Israel) makes me think people will be finding new ones for quite a while.
At least from Finland and using Chrome, following the above leads me to this version, which I've not seen noted elsewhere -- is it specific to Hacker News?
Tried Opera and Safari on the Mac and got a broken plugin graphic with "There does not exist - nor could their ever exist - a plugin capable of displaying this content"
alt text is "Umwelt is the idea that because their senses pick up on different things, different animals in the same ecosystem actually inhabit very different worlds."
I started ready The Little Prince earlier this morning for the first time. The snake + elephant were in the first version of the comic that I saw. My mind was surely the most boggled of any visitor.
It's fantastic, and I enjoyed the different combinations. But I don't really get the one I get, the white disk on black thing. White light at the end of the tunnel? The alt-text doesn't really help me, either.
View source you can see each panel is a different frame and a little note in the source:
title="Umwelt is the idea that because their senses pick up on different things, different animals in the same ecosystem actually live in very different worlds. Everything about you shapes the world you inhabit--from your ideology to your glasses prescription to your web browser."
The JS that's doing the work is at http://imgs.xkcd.com/static/waldo.js. On page load, it sends a GET request to http://umwelt.xkcd.com/story/ghenkEggov8 with params w for the browser width in pixels, h for the height and r with the referrer. The rest of the qualifiers (geoloc and browser (user-agent)) can be determined on the server-side.
As a response, you get back JSON with the alt-text, image URL, and positioning info, which is then used to fetch and position the image.
Fun fact: the server-side is written in Haskell:
http://hackage.haskell.org/package/warp/Well done Randall and davean.