force-kills like sudo reboot will show UI on restart indicating it didn't shut down cleanly, but that isn't reported as a crash. You can see how often you actually crash via about:crashes (and also see what happened)
You should look at about:crashes and see if there's any commonality in the causes, or bugs associated with them (though often bugs won't be associated with the crash if it isn't filed from crash-stats or have the crash signature in the bug)
Google is deciding who are the winners on the web that get better performance (and surprise, google is chief among them, such as the youtube player) - with a very weak justification about why this penalty for security/privacy is worth it (basically "we allow 3PC tracking already, so this can't be worse than that")
I'm down to 7800 tabs (30 windows) on my main desktop. 4K on my other, 2K on my laptops. High was >11000 tabs. Firefox makes this easy; takes ~30 seconds to open all those windows and load a tab per window (plus ~8 pinned tabs)
"% string" in the awesome bar completes against tabs only. Helps a ton, along with some about:config settings that open new tabs from pinned tabs at the right (end) of the tab list. Also windows for separate contexts (though tab groups may also work here now, and avoid the overhead of opening a window/loading a tab at startup).
And "About Tabs" extension from glandium (who works for mozilla)
Many Firefox devs are on Linux; however note that not many Firefox users are on Linux (mostly because desktop Linux has a much much smaller userbase than Windows or Mac). About 4% are Linux, looks like 8-10% Mac, the rest Windows:
https://data.firefox.com/dashboard/hardware
Respectfully, it's a reaction to real comments and opinions that all telemetry is bad, period, and to the continual vocal pushback to virtually any change. No, not everyone says those things, but people do say them. It's also pushback on the idea that there's some nefarious collusion occurring to subvert user privacy; this is a long-term play to try to fundamentally improve user privacy over the status quo, and to give useful arguments and weapons to regulators to crack down on tracking. If you think it can be done better, don't argue here, show up in the standards meetings/issues/repos and make your case. But sitting back and doing nothing is just accepting the crazy level of tracking that occurs now.
Probably the dev needs to use a different method to save (effectively download), much as Adobe Photoshop does when it exports files on Firefox. Not hard to do at all. Likely the same for reading files, if this tool needs that.
OPFS provides high-speed read and write for temporary files and non-exported files; again this is what Photoshop uses. There is a 10GB limit per domain currently. I'm not sure this particular app actually needs that, though
In ~1997ish, the company I was soon to work for licensed Spyglass for use in our Internet-over-cable-TV startup, WorldGate. We ran the browsers in the headend, eventually on custom-designed laptop-chipset-based blades, 10 to a 2U chassis, with 10-20 browser instances running on each blade. (No commercial blades existed back then.) We compressed the screen images and sent them down to settops, with user input via IR keyboards and remotes being sent back up to the headend.
I was hired in Sept 1998 to work on the browser; we had built our own Javascript engine to add to it (since that was kinda required for the web by then). I rewrote all the table code, because it just really didn't work well when you had "too few" horizontal pixels, especially if table widths were expressed in things like %. In the end, after a major redesign of all the table code, it did better than Netscape did in the 'hard' cases.
However, before long, it became apparent with all the additions being made as part of HTML4 that sticking with Spyglass-derived code and trying to update it ourselves to compatibly implement HTML4 (or enough of it) was going to be a herculean effort for a small company (max ~350 people and briefly a $1B valuation (1999), but only around 5 or 10 people max on the browser, including the JS engine.
Given that, I made the decision in late 1999/early 2000 to switch us to the upcoming Mozilla open-source browser, and got deeply involved. The Internet-over-cable-TV part of the company failed (cable companies had other priorities, like breaking TVGuide's patent monopoly, which they paid us to do for them), and we moved onto other markets (hardware videophones) not involving browsers in 2003. I stayed involved peripherally in Mozilla, and when WorldGate dissolved in 2011 I joined Mozilla fulltime to lead the WebRTC effort.
The Spyglass internal architecture seemed at the time to be pretty reasonable compared to what I knew of the NCSA code.
2K? Not bad. My main linux box is running around 7500 tabs in 34 windows right now. Takes about 30s to start, which is fine since that browser gets restarted only every few weeks. My secondary desktop (windows) only has 2750 tabs.
And there are users of Firefox out there with >15000 tabs.
Two reasons for tab hoarding:
1) spatial -- related tabs are close together (frequently open a bunch of related search results; if I come back to them to continue later, they're all together).
2) history -- unlike bookmarks or history entries, tabs retain the forward and back history, so when you return to them you can know how you go there (go back to the search for example).
I do periodically clear out tabs, especially duplicates. The Tab Stats extension by glandium is very handy for tab hoarders
Did you report those sites to https://webcompat.com ?
Also, did you try disabling Enhanced Tracking Protection for those sites? (Shield icon in URLBar)
Try in a fresh profile (via about:profiles)? (Sometimes people tweak a lot of about:config settings, which is generally NOT a good idea - there are no guarantees for those.)
"Linux, for instance, does not support arbitrary call depth and barely uses recursion."
Perhaps there's context missing from that statement? Linux certainly can have arbitrarily deep call depth, depending on the stack size. Are you referring to the kernel? That would be odd, since the paper talks about application code needing to be fully recompiled with retpoline to be safe.
(Of course, that means that all libraries you link with or dynamically load have to be compiled with retpoline too.)
Probably because you have to put a space after % (not sure why, but likely to disambiguate it from other sorts of typing).
It rocks. And I have 1200 tabs open at the moment on Linux64 Nightly (~120 actually loaded). Yes, I'm a tab-hoarder. Bookmarks are useful, but don't give me the easy "I want to come back to this" aspect - once I do close it, I'm done with it; if I bookmark it it's there forever (and the number of bookmarks became unmanagable eventually).
Every so often I do a pass and close out 50 or 100 tabs I no longer care about. Usually I sit at 900-1000; it's time to cull.
Try Help -> Troubleshooting -> Reset Profile (upper right).
That fixes most problems of that sort with (often ancient, migrated since FF 2.0 or 3.0) profiles.
Most things are retained, even cookies. It does remove some things (often things that can cause problems, like badly out-of-date-unmaintained extensions, clears DOM storage, download history, plugin settings):
https://support.mozilla.org/en-US/kb/reset-firefox-easily-fi...
Anant Narayanan has been using a mechanism like this as an example in WebRTC spec discussions - basically a distributed-hash signaling layer (if you will, a self-healing incomplete mesh used to connect nodes for signaling a webrtc call) - and the distributed hash mesh could be a mesh of WebRTC connections with datachannels (no media).
The effect might be an encrypted, hard-to-block (barring deep packet inspection at all the edges or shutting off all access) communication tool. Identity verification would be important for some uses, but not for others.
OOB negotiation via any other channel can be used to join a network (see the article on serverless WebRTC)