9 ms·
I see Wladimir Palant has again commented on µBlock in his comment section [1]: > clear tendency to trade performance for memory use This has been a pattern w
by gorhill 12y ago
I see Wladimir Palant has again commented on µBlock in his comment section [1]:
> clear tendency to trade performance for memory use
This has been a pattern with Wladimir Palant with regard to my work. Outright misrepresentation [2], or criticism through vague statements while never specific enough to understand what part of the code he is referring to.
Nowhere in the code did I trade memory for CPU, both have always been of highest concern. Given that my benchmarks show µBlock does significantly better than ABP on both memory and CPU count, vaguely stating that "clear tendency to trade performance for memory use" is just plain nonsense. [3]
I did ask him to be specific about his statement regarding how I implemented element hiding without injecting gigantic CSS stylesheet in every page and frame, but never received a response, and it is understandable why: he was misrepresenting. [4]
[1] https://palant.de/2014/07/29/which-is-better-adblock-or-adblock-plus#c000761 https://palant.de/2014/07/29/which-is-better-adblock-or-adbl...
[2] https://github.com/gorhill/httpswitchboard/wiki/Adblock-Plus-memory-consumption https://github.com/gorhill/httpswitchboard/wiki/Adblock-Plus... (bottom of the page)
[3] https://github.com/gorhill/uBlock/wiki/%C2%B5Block-vs.-ABP:-efficiency-compared https://github.com/gorhill/uBlock/wiki/%C2%B5Block-vs.-ABP:-...
[4] https://bugzilla.mozilla.org/show_bug.cgi?id=988266#c39 https://bugzilla.mozilla.org/show_bug.cgi?id=988266#c39
Edit: added reference to CPU/mem benchmarks