I've seen one person go thermonuclear trying to prevent ad blocking, and it seriously sucks. Randomly generated content classes and IDs, shifting DOM elements around to avoid them being targeted that way, and later on just completely removing all the divs from the area and manually painting large advertisements with primitives. Made scraping substantially more difficult, but of course not impossible.

One site that still has me beat dynamically loads all of the content and code through a websocket, and various bits are so heavily obfuscated that they look like brainfuck. Finding tools to tamper with SSL websockets is a challenge enough in its own right (Chrome has nicer tools now though).

Interesting. Link to that second site?

For the first one, you mean noscript.net? View source on that to see how they try to push their horrid super registry cleaner pro crap.

are you doing it for the sheer fun of beating the ad-blocker-blocker, or do you need the information off the site?

One is screen scraping, one is attempting to find out how a product actually works. Neither goal has a monetary incentive so it is likely that people have got there before me.

Ad companies want to serve from their domains because they can then correlate traffic across sites and in theory serve up more targeted ads. If every content site served ads directly, that gets more complicated.

And isn't HTTP/2 telling people to stream content from a single server instead of opening multiple connections? I can see HTTP/2 accelerators speeding up traffic by multiplexing requests to the application, CDN, affiliates, etc into a single connection.

It will only multiplex requests to a single host— the layer at which it operates means it will need to connect to both contentprovider.com and ad.networks.com over distinct TCP connections.

Right. The accelerator accepts the connection from the client as a proxy but connects to the myriad backends. The downstream side only sees one multiplexed stream with the responses from the multiple servers.