37 replies
18h11m
When I use Raspberry Pi OS in a Raspberry Pi 4, 8GB of RAM - I feel I already have an excellent, refreshingly stable, late-90s-era experience. It scratches that strange nostalgia itch for that more innocent experience - of early-times WIMP computing.
I can surf the web, edit LibreOffice files, record audio in Audacity on my nice Rode microphone, watch video files in VLC, remotely VNC in, transfer files in and out over SSH's SFTP, etc.
Pretty much all that's really missing, to fill it out, is Zoom (or some such functional equivalent) with a fast-enough frame rate on video calls. And this is not, strictly speaking, the fault of Raspberry Pi, et al.
I used 8GB RAM till recently and I've found it more and more untenable. The primary issue is the browser. Even now on 16GB I restart Firefox every couple of days. But other things also eat RAM like crazy
Running Emacs/Cider I'd have to kill other apps and reboot my REPL a couple times a day. Emacs would also leak memory and need a restart every couple of days
I guess the primary issue is the chair-->keyboard interface. The fact that browsers can open unlimited number of tabs doesn't mean you don't have to do a little housekeeping.
Plenty of people still use systems with 4GB and lower and it works fine as long as the number of tabs they open is limited.
The primary issue is that browser developers are people that can afford kitted out Macbook pros so the system isn't designed to scale to small/weak systems :))
I don't believe a browser couldn't be designed to have a small RAM footprint. All my tabs could be suspended and saved to disk when in the background (and not spinning any tasks). They can be read back into RAM near instantly when I tab back to them
No, Firefox and/or raspberryos just not well engineered for this usecase.
My Chromebook with 8Gb ram has dozens of tabs and web apps open in Chrome, runs one VM with Android and another VM with Linux in turn running Firefox and more. All without breaking a sweat.
I run Firefox on 8 GB without any trouble whatsoever. But I also rarely open more than a dozen tabs.
I use Firefox on an 8GB, early 2013 MBP, with hundreds of tabs and an extension, AutoTabDiscard that unloads/suspends them after a couple of hours.
Works beautifully. I have to restart the computer about once a month because of Catalina bugs, but Firefox is super stable.
I shut down my laptop at the end of the day and turn it back on the day after, regardless of bugs. Why do you try to reboot it as little as possible?
It’s my home computer. It’s there to be used intermittently when needed at random times. It sleeps drawing almost zero power. Why should I shut it down?
Shutdown takes 20 seconds. Startup requires the FileVault password, then 20-30 seconds, then a login, then another 20-30 seconds until desktop is usable (and a few more until Firefox is).
If this was my work computer, it wouldn’t be so inconvenient to restart / shutdown once a day. But for what reason?
I guess that the main difference then is that I have to wait less to boot my machine. Did you consider hibernation? It would be a bit slower than sand-by, but then it would draw exactly zero power.
The startup/shutdown time is possibly explained by being an 2013 machine, but I have little reason to replace it right now - it’s only 8GB, old slow CPU (by modern standards), old slow SSD - but it does Firefox, thunderbird, the occasional Python script and a few more things perfectly well.
I’ll replace it when it breaks.
With respect to power draw - there is no simple way to force hibernation on Catalina AFAIK, but the power draw in sleep is minuscule - it hardly registers on my wattmeter (and e.g. it loses only 2-3% percent per day of battery while sleeping).
I mean nominally that sounds like his preferred experience?
I'm similar, I prefer maintaining state with things until I'm done with them, which makes the current models so frustrating, for all Apple talked about skeuomorphism, for me it's always felt so fake, it's only ever skin deep, I open a webpage and until I'm done with it, it should stay that way.
I've navigated down a third of the page? I've partially filled in a form field? Keep it! There's probably a reason I put that there!
It's not like when I put a piece of paper down on my desk it resets to it's original appearance and orientation every morning. It retains the scribbles and notes! Maybe you like someone else tidying your desk every morning, but I hate it!
Real things exist, our memories exploit these properties so well and what do we get, software that's all about returning to some pristine state that makes it harder for me to recall and use.
I'm curious if they're going to do this same thing with their spatial os, or whether they'll work out that persisting things until people are done with them is a feature.
I wasn't complaining about someone else's experience, I was just curious to see why he preferred it that way.
Sometimes I also use stand-by or, if it is to keep the state up until the next day, hibernation. But that's rare. In most cases, I just use Firefox's function to restore my previous browsing session. That would not restore half-filled forms, but I rarely deal with forms, especially long ones. As for reading or editing documents, most software will open the document at the point you where when you last closed it.
i wish firefox had more granularity with its tab unloading feature. one method is restarting browser. but by default it’ll unload tabs if you’re low on memory. you can force em in about:unloads (hint- about:about if you ever forget)
i wrote a simple firefox extension that unloads em all with a button click, but it’s no different than restarting ff or spamming clicks in :unloads
We already know the solution. We run the browser in Wasm inside the browser. That way it only has one tab open and that tab is doing 99% of its work inside of a wasm env. I thought firefox would ship new versions of firefox inside of firefox at some point. I guess that point is still in the future.
This of course is how electron should work as well. A canvas only frame that loads whatever rendering system you want, which could be a browser, or it could be Unreal engine.
I can't tell if this is sarcasm or not, but assuming it's not...
How many levels of abstraction do we need to run software reliably? The fact that browsers have effectively become operating systems should be worrying enough.
No wonder we have all these fantasy projects that take us back to when our computing environments were actually pleasant to use. I would partly blame the invasive tracking and needless complexity of modern OSs for that, but the ever growing software layers around hardware makes no sense at all. We should question any such design decision, and strive to simplify this ball of complexity instead.
Isn't this what SDL (https://www.libsdl.org/) is for? Some cross-platform (and pretty light) hardware abstraction where you can have a canvas, do 3D, audio, whatever...
There’s AutoTabDiscard which lets you set a timeout and other rules for when to discard.
Is AutoTabDiscard still helpful/necessary with whatever automatic stuff Chrome/Firefox does nowadays? I run a lot of tabs (100s-1000s) and this sounds potentially very helpful if it does something beyond what the browser automatically does.
I think Firefox will delay loading on session resume (that is, if you restart Firefox, tabs will not load until you actually switch to them),but will not unload anything automatically - so if you open 200 tabs between restarts and keep them open (which I sometimes do, for a week or two until I close them) it makes a difference.
Try Sidebery extension.
'Auto Tab Discard' was a game changer for me using firefox. I was about to upgrade to a 64gb laptop.
it helps but strangely doesn't eliminate the problem entirely
Perhaps I don't see it on 32gb but can relate with cider/emacs.
Really? I can only hold one firefox session at a time in 8GB of RAM, but I've always assumed that's because I keep an unreasonable number of active+background tabs open.
12GiB and copious swap. 4 profiles open with 50-100 tabs loaded at any time.
The only problem is the accumulation of CPU use from web apps.
Consider adding more swap space so that older tabs have an out-of-the-way place to stay.
NoScript and only turn on JS when required, and only for the site itself. Hugely reduces memory requirements
Not even close, if you told that about a ESP32 based system then I would have agreed.
My 2003 bought multimedia Athlon XP desktop had 512 MB!
The PC I was using in the late 90s had probably 32 MB of RAM after an upgrade... when I built a PC (2001?) with 512 MB it was looking like an infinite amount of RAM...
I remember just laughing when I heard that Adobe had fixed a bug that occurred when running Photoshop in more than 1GB RAM on Mac OS 9. It seemed like such a theoretical thing to have that much memory.
Zoom alone takes 2.4GB of RAM, just after being launched and starting a meeting - and no-one's even joined the meeting yet.
Not even close.
Moat of us had about 16 or 32mb at the time IIRC
Unless you had RAMDoubler(TM)! Remember that?!
I started my studies on 2003 and highest I could do was 768MB of RAM. I remember this amount exactly, because my LL(1) grammar compiler was leaking memory on my pre-presentation test, and I had to present it to pass the course. Every MB of memory counted, and it started swapping. I was praying to make it pass but when the amount of swap increased above the amount of RAM I gave up. However I was last in the queue to present, and the teacher told me he needs to go, as it was too late. I was so happy I barely could hold my laugh. Came to my dorm and fixed it next day. Core memory. That's how I also fell in love with my wife :D She was doing the extreme programming with me
You weren't born in that late 90's era right? :D
Correct. 70's
While all those pointing out that 8GB of RAM was mainframe stuff in the 90's are absolutely correct, I would offer that the software bloat between 90's software and modern software does make the _experience_ roughly comparable.
Except for the built-in HDMI video and the seamless plug-n-play networking that is...