Max Nikulin [2025-02-13 09:04:38] wrote:
> On 11/02/2025 06:06, Stefan Monnier wrote:
>>>> Web browsers are bloated monsters.
>>> Does anybody have a bookmark for an article with a balanced view on degree
>>> of responsibility of
>>> - web site developers,
>>> - users,
>>> - browser developers
>>> in respect to browser resource consumption?
>> Reminds me a "recent" rant of mine:
>> https://oldbytes.space/deck/@monnier/113290809168999071
> Thanks. I hope, somebody has wrote a more detailed article in this
> spirit. The idea is to adjust user expectations and suggest a way to use
> browsers when system resources are scarce.
> By the way, chromium may display memory footprint in tooltips for tabs. In
> some cases it is more prominent than about:processes in Firefox.
But that still points to the resources used by the browser, whereas my
rant above is above displaying the resource requirements of the webpage
itself since I think.....
>> When you view simple webpages,
>> Firefox is clearly bloated (I count 8 processes for a vanilla `firefox`
>> visiting my trivial last-century-style home page at
>> https://www-labs.iro.umontreal.ca/~monnier/, with an RSS of at least
>> 100MB).
> This page is a kind of a corner case nowadays. Notice that the mastodon link
> above is useless without _javascript_. I think most of users prefer faster
> reaction when they open webmail, video players, word processors, etc.
Right, these are web applications, not web pages. There are still loads
and loads of web *pages* which *should* be like the "coroner case" but
instead are implemented as if they were web applications, loading and
running crazy amounts of _javascript_ code (mostly for nefarious purposes
such as propaganda and associated user tracking)......
..... and then.....
Knobs in the browser can't help because (almost) all the logic that
controls resource usage (above the base 100MB) is in the downloaded
_javascript_ code. By the good ol' halting problem, the browser can't in
general know which piece of that _javascript_ code will control what.
And I really hate uncontrolled non-determinism
:-)
Let's assume for now that the browser indeed has no influence over _javascript_ runtime behavior. We have plenty of infrastructure to control resource consumption by any process but almost no one uses it apparently: cgroups.