Slashdot Mirror
Raspberry Pi Gets a Brand New Browser
sfcrazy writes The Raspberry Pi team has announced a new browser for Raspberry Pi. They had worked with Collabora to create an HTML5-capable, modern browser for Pi users. While announcing the new browser, Eben Upton said, "Eight months and a lot of hard work later, we're finally ready. Epiphany on Pi is now a plausible alternative to a desktop browser for all but the most JavaScript-heavy sites."
No, it has nothing to do with the WWW. It has a lot to do with poorly designed web browsers written in C/C++ which leak memory like a sieve and treat memory like everything has unlimited quantities of it. A well designed web browser supporting ALL of the HTML, Javascript and other web standards could use a small fraction of what Firefox uses if it is written properly and in a better language like Ruby. With some quality programming, the web browsers could use a fraction of the CPU power they do now by storing off screen graphics in compressed formats, for instance, especially since ending up with disk caching due to using up the RAM is far more resource intensive than decompressing images. There is no reason, NO REASON that Firefox should eat up 1 GB of RAM. The Javascript and HTML engines supporting ALL of the latest standards and ALL legacy standards only take up a few MB of RAM. There is massive memory leakage going on these browsers of forgotten image and multimedia data. Observing the behaviour of Firefox it cannot be anything else, when you open 20 windows and then close all except the first one, memory never goes back down. If web browsers were written in say, Perl or Ruby, I gaurantee that you could easily have browsers that are many times faster than current browsers because the memory usage and swapping would be much less, plus would be much safer without all of the buffer overruns.
Had doubts wether trolling or not, but you got me at "perl".
So because they insisted on using a crappy 12-year-old design ARM11 CPU, they need a custom browser to compensate.
Why not make a Raspberry Pi model C, with a Cortex based CPU? If they used a modern A17 at 1.4Ghz, it would be just as low-power and have ~8x more performance.
Because they like to sell their the Pi's at or below $35?
Be seeing you...
Not at the first mention of Ruby?
Perl Programmer for hire
There are weird ruby lovers out there.
There is no reason, NO REASON that Firefox should eat up 1 GB of RAM.
Firefox caches images and rendered pages so that things happen reasonably fast. A 1000x1000 colur image at 24 bit is already 3M. There is no way a modern page with an image or two would fit in a few M.
Once you have 800 tabs loaded (something I tend to do) all running JS and using images all over the place, memory vanishes fast.
There are basically several problems. Yeah firefox isn't perfect and could do with some optimization. But, many websites do pointless bloaty stuff and require scads of JS and huge images just to show some text. And then there's the browsing habits.
These days I run with NoScript, and a tab unloader, and I'm picky what scripts I allow. It makes firefox usable on my 1GB netbook. It's taking up 176M at the moment. Actually it's got a bit better recently. Even with those it needed restarting to clear the ram, but now it runs substantially longer.
SJW n. One who posts facts.
Let me get this straight.
- You suggest Ruby, which is known to be notoriously slow, would be a good choice of language to write a high-performance browser
- You suggest burning CPU cycles for compression/decompression helps save CPU cycles
- You have problems with the idea of a browser written in C/C++, but subsequently suggest using a language written in C/C++.
Either you're trolling or are so thick that you don't see your own doublethink.
Maybe browsers aren't as efficient as they could be, but current "web design" practices are definitely to blame for the general slowness. For example, this page (Slashdot Classic) wants to load (mostly scripts) from eight additional domains. I have a couple of web sites of my own, and they all load instantly and scroll without a hitch, even though I use scripts to enhance the presentation too, but of course I learned web design and programming in the 90s, when computers with 200MHz CPUs were still around. Pages these days load scripts to show you social networking buttons (and track you), load scripts to show you adverts (and track you), load scripts to drop referral cookies (which help track you), load scripts to show you maps (and track you), load script libraries and scripts to show you pictures (and track you) and load scripts from at least three "analytics" services to track you. Pages like that take several seconds to load, and after they're done loading they're still slow, even on a multi-core, multi-GHz CPU. The web is FUBAR. I wouldn't use it without AdblockPlus. Major ad networks are blocked at the DNS level. Lately I've added the RequestPolicy addon to my defenses.
It's a good thing.
Code of this lighter, faster browser for the Pi may end up in mainstream browsers, which will then be able to handle heavier, more complex, and more poorly-designed websites, which will in turn lead to an even faster, more efficient browser for the Pi.
The result after all this effort is that the Web will still be at least as slow as before, just the way we grumpy old gits like it.
I'm actually using Epiphany, the new browser, to post this. Slashdot was one of the first sites I visited and co-incidentally there was an article about it right at the top! So far, it does seem to be a nice upgrade to the previous Midori browser, which I found essentially unusable.
1) Ever heard of caching? 2) Browsers do not need to support HTML standards, but real-world HTML practices, which is messy. Such as tested by Acid2.
NB: The message above might reflect my opinion right now, but not necessarily tomorrow or next year.
Agree. I think there should be something between Intel NUC and R-Pi. I think R-Pi has merit in that its just enough compute and memory to do almost everything, but I think in 2014 that more can be done with $35-$50.
Legalize the constitution. Think for yourself question authority.
Like this MIPS attempt ? http://www.pcworld.com/article...
Or any other or wannabe rpi competitors.
http://validator.w3.org/check?uri=http%3A%2F%2Fwww.slashdot.org Errors found while checking this document as HTML5!
which leak memory like a sieve and treat memory like everything has unlimited quantities of it.
You have no idea what a memory leak is. None. In fact, I don't think I've ever heard a layman use the term correctly. A memory leak would lead to the application using a hell of a lot more memory than Firefox ever could in just a few seconds, if not immediately, and lead to the application faulting. And, for a point of fact, modern computers start with 8gig of memory, and then there's the page file/swap space. Effectively memory is unlimited, at least in comparison to the paltry 1 gig of memory Firefox is using. Firefox is a browser meant for browsing... and if that's what you're doing with it, that 1gig of memory is nothing. What background apps are using up the other 7gig? If you're running it in the background to do something, you're using the wrong application. Pandora has its own lightweight app for example.
A well designed web browser supporting ALL of the HTML, Javascript and other web standards could use a small fraction of what Firefox uses if it is written properly and in a better language like Ruby.
Example? Why aren't we all using it? Firefoxes massive advertising budget?
There is no reason, NO REASON that Firefox should eat up 1 GB of RAM.
Of course there is. You seem to have this idea that memory is the be all end all of resources. This was true in the 70s and 80s. Memory was $$$ then and you gave up everything to save memory. But now memory is insanely cheap... so you can use more memory if it decreases load times, lowers CPU load... or in the case of firefox, if it reduces badnwidth usage. Now-a-days bandwidth is the $$$. Many people in the world are on capped internet connections or they get charged by the gig downloaded. So Firefox has made caching its #1 priority, You are saving money on you monthly internet bill at the expense of storing that data in memory.
800 tabs. Hyperbole much but the question is how much. I guess maybe 80 tabs ?
Spill it, I need to know. What if I show you mine first: I'm currently at 17.
How bad is your tab hoarding ?
Bullshit. Time is no factor in the mere question of whether something leaks memory or not. If a program fails to properly release memory then it's a memory leak, whether it's one gigabyte per second, or one byte per day.
What about the Banana Pi http://www.bananapi.org/p/prod....
Considering you can buy a 7" tablet with a 1.5GHz dual core Cortex A7 with 512MB ram and 8GB flash for $38, I don't see why the Raspberry Pi, with no flash, no battery and no screen couldn't be even cheaper than $35 with a different SOC.
Apparently because the A20 boards are ~$50 they're not comparable to a Raspberry Pi, despite the Raspberry Pi being completely useless without an SD card that isn't included in the $35 because it has no onboard flash.
"Even with those it needed restarting to clear the ram"
The first step is admitting you have a problem.
That's not an answer. That just says they made a stupid choice. Face it. Anybody with half a brain buys the BeagleBone anyway. What's an extra $10-20? One or two pizzas.
I blame people who have no idea that sending a 1.8MB, 3000x2000 pixels PNG is a bad idea both for the size of the file and the RAM required to store the file and the RAM required to decompress the image in order to display it. And that's only the background image of a single page.
Get free satoshi (Bitcoin) and Dogecoins
Made in China vs made in UK?
What's an extra $10-20? One or two pizzas.
Maybe I'd rather have the pizzas than waste money on something that I don't need?
How can we continue to believe in a just universe and freedom to eat crackers if we have no ale?
Well I bought 3 things from Aliexpress:
A 32gb SD card which craps out if you copy more than 2gb on it.
A set of 3w garden lights which draw just over 1.5w at full power and burn out if you even slightly raise the voltage.
A Bluetooth speaker with the beats logo which has approximately 10 min battery life.
So yes you can make anything for any price, but I'd rather buy something which works rather than something unrealistically cheap.
What's an extra $10-20?
More than half the price of another raspberry pi.
People with brains buy a system to suit their needs. People with really good brains optimise that system to get maximum performance. Everybody else needlessly spends money on something more powerful than they need.
I have 19 open, organized by task (the last 3 are just general browsing, most of the rest are references for a project I'm working on). Related things are close together. 18 of the 19 are currently visible in the bar, with the titles legible. It's a little more convenient than having them down in the taskbar would be. If I have multiple activities that each need a large number of tabs, that's why God invented tab groups.
It is pitch black. You are likely to be eaten by a grue.
At least the first 10,000 Raspberry Pi's were made in China.
They didn't even start manufacturing in UK until September 2012.
They were still manufacturing in China in December 2012.
I'm afraid it is you that are wrong. A memory leak just means that it is not properly cleaning up after itself i.e. releasing memory when no longer required. It does not mean that the app needs to crash. It does not mean that the app needs to keep leaking until all memory is used and it crashes for it to be a leaky app. For example clicking on the back button or some other control may cause a few bytes to be consumed that are never released because of poor programming/memory handling. This is a leak. Will the app crash? On a modern machine it is not likely as you'd have to click a lot of times in order to consume enough memory. However it does not stop it being a memory leak.
Correct. The GP was wrong.
In the case of Firefox start with one tab. Check the memory usage. Open lots of different tabs, for instance, by opening links from your favourite iGoogle style dashboard. Note the memory usage. Leave them sit there for a while. Then go through closing them down. I guarantee you that you will not get back to anywhere near the starting level. That is a memory leak by definition.
No, it is not. You gave the correct definition earlier: the app is not releasing memory which was allocated but is no longer necessary.
Without knowing how Firefox works internally it's difficult to say that closing tabs should recover that memory, but I'm going on a limb and say that it shouldn't. For one thing, you can reopen closed tabs, and they'll come back essentially as they were left, including the ability to navigate using the back button. Where do you think this information is stored?
There are also other complex mechanisms involved in showing a page that might have an impact on memory even after tabs are closed. Image and other caches may be kept in memory for some time rather than being dumped to disk straight away; the JS engine may keep components loaded for faster access on subsequent runs; if you open a page with flash, the plugin will probably remain loaded even if you close all pages with flash content; etc.
A modern browser is a lot more than a HTML renderer. It's a complete platform, and honestly 1GB of RAM after heavy usage doesn't seem like all that much considering all it does.
What's an extra $10-20? One or two pizzas.
(a) For me, $10-$20 is no big deal, especially since I don't need many small board computers. What about a school district that is buying Pi's in bulk? A very poor school district?
If the BeagleBone is a better deal for you, then buy that. I plan to buy that (BeagleBone Black, thank you very much). But I don't think the RPi guys are "stupid" or "made a stupid choice".
(b) If you look at the history of how the RPi got made, it was always going to be a Broadcom SOC with that particular CPU core. The RPi guys didn't have total freedom to do any damn thing they wanted to do; the head RPi guy worked at Broadcom.
I for one am glad the RPi got made. And people are doing lots of cool stuff with them. At the price, they are being used for things that you might have used an Arduino for, because why not.
So don't moan about how the RPi isn't ideal for a desktop replacement. It has its place and it's not a "stupid" thing at all.
With Seamonkey you can still rightclick an image and select 'Block images from url.com' and bip! Sorry, web 'designer', you failed to impress. Firefox has gone over to the darkside and doesn't have this feature. It's nice to have the images from adservers blocked so the aren't even loaded.
Do you have them open because "lulz look how many tabs I can have open"? Serious question. Because there must be tabs there you haven't looked at in months, and if they disappeared you'd never know.
modern computers start with 8gig of memory
Is the ASUS Transformer Book, a 10" convertible laptop computer, not "modern" because it ships with only 2 GB of RAM?
Besides, not all computers still in use are modern. I do most of my web browsing on a four-year-old Dell Inspiron mini 1012, a 10" laptop that came with only 1 GB of RAM and runs Xubuntu. Flashblock helps keeps Firefox below half a GB, after which point the bottleneck is not memory but the fact that Firefox uses only half of the CPU. Though an Atom has two-way simultaneous multithreading, Firefox is still single-threaded which brings a longer wait for complicated JavaScript and CSS to finish processing, especially on things like Cracked.com or Slashdot beta. My first-generation ASUS Nexus 7 tablet computer is stuck at 1 GB as well.
and then there's the page file/swap space.
A lot of computers without a rotating hard drive, such as my Nexus 7, can't afford to swap. Instead, they have a harsher OOM killer.
Firefox is a browser meant for browsing... and if that's what you're doing with it, that 1gig of memory is nothing. What background apps are using up the other 7gig?
I'm guessing that when no application is using part of the RAM, the chipset could power down unused RAM to prolong battery life.
It has its place and it's not a "stupid" thing at all.
Keep in mind that this is Slashdot. As long as a product doesn't meet that autistic need for every little factor to be 100% compliant with one's personal tastes, it's a stupid waste of time.
While the rest of your comment may be sarcastic, this is actually a very good idea. First, it is actually many, many times faster to recompute a value than to pull it from memory:
Secondly, even if we're talking about content that isn't already available in compressed form (which most web content is), there are simple compression methods like LZOP, where the extra CPU overhead is still faster than the increased amount of I/O needed to transfer the equivalent amount of uncompressed data.
"lzop is usually IO-bound and not CPU-bound"
http://www.lzop.org/
Frankly, I'd be happy to turn off all Firefox caching if I could... I never open more than 20 tabs, and certainly not with huge images, yet on my 1GB system, Firefox starts causing swapping to disk after every 8 hours or so, and needs to be restarted, and I certainly don't buy the cop-out that my add-ons (Adblock and NoScript) are to blame, while Firefox would be perfect otherwise. And this is with the LTS/ESR version, which is supposed to be the super-stable version.
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant
Web pages don't usually have 1000x1000 images, with the exception of wallpaper sites, and people sure as hell don't keep 300 tabs open with those images, to justify Firefox using 1GB+ of memory and swapping to disk... Never mind the fact that Firefox could cache the compressed version of the images, and re-render that part of the page when the tab gains focus.
If we're going by Firefox, it seems there's no way a modern page or two will fit in a few GB.
Yes they do, but Firefox is smart enough to delay the loading of tabs until they gain focus, and could STOP the processing of all JS and unload decompressed images from cache when they lose focus again.
My browsing habits sure aren't to blame... I use NoScript and Adblock, too, and never have more than 20 tabs, and yet about every 8 hours I need to close Firefox because my 1GB of RAM is exhausted, and the swapping to disk is making my system unresponsive... This with the lightweight Fluxbox as my WM, and absolutely nothing else running.
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant
On such example:
http://alanwalkerart.com/wp/wp...
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant
But he's clearly smarter than everybody who spent years thinking of and designing an enormously successful product.
Nah, he's really a fuckwit but thanks to Dunning-Kruger he doesn't know it.
Hail Eris, full of mischief...
E pluribus sanguinem
Rasome. (As in, Jhon Carter movie)
You know, generally speaking, you can get back to a website even after you close a tab. You shouldn't be afraid of losing that page.
It's no longer called "Epiphany". In what seems like an epiphany, the GNOME developers decided that it's much, much, easier to search for help for a browser called "Web". Great idea, there, guys. Was this intentional, to prevent intelligible bug reports from less sophisticated users?
One wonders whether they actually "eat their own dog food", or if they do, if they understand that the average user of GNOME isn't a GNOME developer.
At first I thought, "oh great. Fucking ruby nutball."
Then I got to "perl". Well done. Well fucking done.
FireFox thus far have refused to implement a configuration feature where they themselves limit the amount of memory they use. They say it's already built in and auto-tuning based on the amount of memory the OS reports. It's about time that FireFox stop being so arrogant and just let me set a limit, because I don't want them to eat all memory that I want to use for other applications that now have to resort to swap because a browser eats over 2G of my ram.
I was promised a flying car. Where is my flying car?
With firefox + tree style tabs it's perfectly work-withable to have dozens, if not hundreds of tabs open. You can easily see at least 40 tabs on the screen at once + collapse trees that you don't need at the moment.
People and programmers have been spoilt by multi-GHz multi-core CPUs. People used to edit video, design space ships, simulate physics, ray trace liquid metal and just about everything else on far weaker machines. It good to see that some people can achieve good performance on limited hardware. The raspberrypi foundation are funding work all over the free software stack, which will benifit plenty of people who have never seen or used a pi.
How do you manage that: several windows dedicated to different purposes, a grouper plugin ?
I recently stopped using AdblockPlus, for Ghostery+NoScript+Flashblock already block everything ABP would block otherwise
CLI paste? paste.pr0.tips!
843 today, over about 20 windows ... each window to different purpose, yes. Session Manager to save them reliably. Mostly it's documentation and stuff-to-read relating to various projects in parallel. Firefox's tab handling is awesome compared to e.g. Chrome ...
Hehe, I first read "Microsoft server" and wondered WTF...
Everything I write is lies, read between the lines.
I have in the high thousands of bookmarks, that's another thing FF shines at with the bookmarks sidebar and search, Chrome sucks at finding bookmarks. But it's much easier and quicker to open a bunch of tabs (right-click), keep them open and topically-grouped in one window while needed, and close them after, than bookmarking things which might never be returned to or might not turn out ot be useful, and then trying to find them.
A memory leak would lead to the application using a hell of a lot more memory than Firefox ever could in just a few seconds, if not immediately,
[citation needed]
It is perfectly possible to have a slow memory leak. Saying otherwise is dumb. If the leak is in an infrequently-called routine, or simply does not occur every time the leaky routine is called, then it won't happen the way you imagine that it will.
But now memory is insanely cheap...
...except on mobile platforms, where the power budget has to be considered along with the cost of the memory. And did I mention that DDR4 costs about twice as much as DDR3?
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Having used both the Beaglebone and a Pi, I actually prefer the Pi. There is just more developer support and thinks work better. The BBB has a bunch of annoyances, like Angstrom Linux (blows), and that the on-board memory is just too damn small (2GB is not enough). Booting off of external SD works, but you have to go through this annoying contortion of holding down a button while plugging it in to make it work. Even once you get Debian booted off of the SD, it is missing stuff that the Pi has, like Google Chrome. Once you use the factory overclock option on the Pi, it is just as fast or faster too. In my benchmarks of a somewhat weird application I wrote the Pi with the standard OC (scale up to 1Ghz) was actually slightly faster than the BBB. I would overclock the BBB to make it more fair, but again there isn't the developer support and I didn't find a way to do it.
Add in that the BBB is like 60% more expensive on top of that and it's just not a good buy. If you need a bunch of PWM channels or it's far superior general I/O capabilities then yeah go for the BBB, but if you're looking for a small Linux box for more general use then the Pi is better.
I read the internet for the articles.
Recent versions of the BeagleBone Black have 4 GB of internal flash and come with Debian pre-installed.
That being said, I have a RPi, mainly for use as a media center (one of the best uses for it).
Oh, no! You have walked into the slavering fangs of a lurking grue!
That doesn't necessarily mean that Firefox is leaking memory. It could mean that it is keeping it for its own reuse rather than releasing it back to the system. If the expectation is that the browser user will soon need the memory again (perhaps by opening 19 new tabs), that is likely to be more efficient than releasing it to the system pool and then later asking for it again. On the other hand, if the user next needs all that memory to run something OTHER than Firefox it's not so good.
The Chrome model of using a separate process for each tab leads more naturally to release of memory to the system, as there is no alternative when the tab's process is terminated.
Not caching uncompressed graphics, as you are suggesting, is an interesting possibility for memory-limited systems. To see whether it is a net gain it would be necessary to build a browser that did it, preferably by doing a modification of an existing browser like Firefox (so you could compare apples to apples) rather than creating a completely new one.
Memory may be cheap, though not as cheap as it was a couple of years ago; the street price of RAM has nearly doubled since then. But lots of low-end laptops have rather low limits on how much RAM you can install, like Bay Trail laptops with 4GB, and mobile devices like phones and tablets can't be upgraded at all and are likely to have a measly 2GB.
--Submitting personal experience, Palemoon on Linux 64-bit is beating Firefox all hollow WITH THE SAME PLUGINS running.
--In my work environment, Xubuntu 64-bit, Firefox would regularly use ~2GB+ of my 6GB of laptop RAM - and become extremely slow. I open and close tabs all night. Palemoon is *much* more memory efficient - and I haven't noticed the same slowdown effect.
--Right now I have 15 Palemoon windows open and God knows how many tabs, but one of the best changes they made to the Firefox base code is not loading all tabs on browser launch - only when they're active. That said, Palemoon is the better browser in my experience.
.
== WolfriderV6 == I'm willing to admit that *I just might* be wrong... Are you??
It's a bit more complicated than just decompressing the images. I suspect the browsers are keeping the complete HTML rendering of as many open tabs as possible so you can switch to them instantly. They are also keeping any Javascript code on the pages active (if not actually running) so they will be in the same state as when you left them. If each tab had to be re-rendered when you switched to it there would be a noticeable lag when you switch, and users wouldn't like that.
Sorry, but web pages get rendered into images before displaying them. (Though at least Firefox's semi-recent versions don't bother rendering web pages until needed when you crash&restart Firefox, which I do all the time - usually not on purpose, though I'll occasionally do it to scavenge memory or when performance has become unbearably slow.)
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
I don't know how many tabs I have open right now, probably around 500. And while most of those are mostly text, Firefox might very well keep them as full-color images to avoid re-rendering when it needs to display them.
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
Anyone else having trouble with it being slower than an infected IE running on a 386? Fast doesn't come to mind with a two-minute initial startup time and minute long page loads for Google searches. I'm running raspbian with a medium overclock.
noi that phong thi nghiem, bÃn thà nghiám