News for hackers. Stuff that matters.
Re:Performance Impact of Ad Blockers (2020) by floatingatoll on Sunday February 07, 2021 @11:23AM (#26055992)
Running a persistent content blocker as a third-party HTTP proxy process is more efficient than running in-page: you get all the CPU benefits of content blocking, without the RAM drawbacks of loading JS blocklists through page injections. I would not consider the browser extension model to be viable for content blockers in light of these statistics, as the overhead of a cold start simply can’t match the readiness of a network proxy.
Re:Performance Impact of Ad Blockers (2020) by chias on Sunday February 07, 2021 @12:38PM (#26056534)
Yet here's a drawback: you can't handle cases where the domain doubles both a service that you depend on and as a source of advertising and tracking that you wish to block. Facebook and Google are the two that come most readily to mind, but there are countless others.

EDIT: my apologies, I misread your comment and assumed you meant host-based blocking. You seem to be advocating for the proxy to parse the page contents and do the ad-blocking there, before serving it on to you. It's an interesting idea. At scale it would have some significant hurdles to overcome WRT TLS, but I am nonetheless intrigued.

Re:Performance Impact of Ad Blockers (2020) by userbinator on Sunday February 07, 2021 @05:12PM (#26058721)
At scale it would have some significant hurdles to overcome WRT TLS

Proxomitron and other MITM proxies have been doing this for years --- you do need to generate your own root CA and change the TLS config of whatever clients will go through it so that it trusts your root, but after that it works perfectly. As a bonus, you can do "HTTPS upgrade" with applications that don't even support HTTPS or only old TLS versions.

Re:Performance Impact of Ad Blockers (2020) by chias on Sunday February 07, 2021 @07:04PM (#26059406)
Those aren't third-party HTTP proxies, they're local HTTP proxies.

If you want to do it locally, then you're paying the CPU / RAM cost yourself, not to mention doubling a lot of the work (the page needs to be parsed / interpreted / JS executed by the proxy, and then again by your browser) -- so why not just do it in the browser?

If you want to do it third-party, you need to trust this third-party because they're terminating the TLS connection for you (not to mention being fundamentally incompatible with TLS 1.3)

Re:Performance Impact of Ad Blockers (2020) by nick_ on Sunday February 07, 2021 @04:43PM (#26058479)
AdGuard has this ability. It will even install a certificate to your OS so it can parse and block TLS connections (but only if you explicitly instruct it to).
Re:Performance Impact of Ad Blockers (2020) by floatingatoll on Sunday February 07, 2021 @01:49PM (#26057093)
Pi-hole, for example.
Re:Performance Impact of Ad Blockers (2020) by chias on Sunday February 07, 2021 @07:06PM (#26059418)
Pi-hole is not a proxy. It is a DNS resolver that is configured to fail resolution to given domains.
Re:Performance Impact of Ad Blockers (2020) by dgllghr on Sunday February 07, 2021 @03:08PM (#26057779)
I believe Pi-hole is DNS only