Facebook adds 5 divs, 9 spans and 30 CSS classes to every post in the timeline

I wonder at which point will we just stream pages as interactive video content with a text optimized codec. Most pages on the internet are bloated to the point where it might just make sense. (I am not advocating, I am predicting)

At a previous job in the not-too-distant past, we needed to support IE7 for contractual reasons. I'd unironically floated the idea of launching a VM on the server running Firefox, and starting a VNC session, streamed over frames of a GIF. Clicks and keypresses would be transmitted with AJAX calls.

In a later project, we played with the idea of using an asm.js-compiled Webkit to render SVG (with embedded HTML) into a canvas due to mixed browser support.

Long story short, it's not inconceivable.

Write once, run everywhere promise of the web delivered like never before.

There was a Chinese mobile browser that essentially did this if I recall. And didn't Amazon have something like this for one of their earlier Kindles?

That’s what Flash was all about back in the day, more or less.

This is pretty crazy. I love it.

Deep learning is already able to find the ads even within video content, so the cat-and-mouse game is still on :)

Of course locking down the computer helps.

Doesn't citrix solve this problem?

Facebook should try to adapt screamer-prank tech[0] for ads. Even if an ad missed your attention, you'd still have to pay 1 laundry bill and 1 bill for stuttering-aid couching.

Jokes aside, even with rendering to canvas, ad can be blocked with image recognition. Long time ago I used Sikuli Script for UI testing automation, Sikuli recognized parts of screen and worked very fast; combine it with an app like CinemaDrape and voila.

[0] (loud noises warning) https://youtu.be/469zNXTCHdk?t=10

Then we would finally have implemented the interactive TV envisioned in old futurism. And the web would be dead.

If you stream movies from sites like 123movies, they play a similar obfuscation game but are moving towards your hypothetical.

While many players, despite deeply nested html, still point directly to where the content is hosted (ex, a file on openload.co), an increasing number take your mentioned approach and load the content as a blob so folks won't bypass their ads.

Sounds a bit like Apache Guacamole, which is sometimes used for presenting old software to users via a web setup.

I worked on a project a few years ago to render screenshots to imagemaps, then used server side imagemap processing to click.

Made modern websites work in Netscape 2 :)

IIRC you can directly download a blob from the blob internals page; I think you can also see them somewhere in the developer console, and I've got a bookmarklet that allows feeding a blob URI into it for download.