Arch Philosophy/Structure Applied to a Browser

thisllub · 2009-04-20 02:34:39

It seems to me that the biggest flaw of Firefox is extreme memory usage when large numbers of pages are opened.
This could reasonably be solved if it implemented disk based persistence for non active tabs.

leo2501 · 2009-04-20 02:44:49

thisllub wrote:

It seems to me that the biggest flaw of Firefox is extreme memory usage when large numbers of pages are opened.
This could reasonably be solved if it implemented disk based persistence for non active tabs.

It's SLOW... lol, in my intel atom 1.6 it lags several seconds to switch between tabs or open a page in a new tab, even gmail is painfully slow... with links -g for example, browsing the internet is blazingly fast, something is slowing it down under the hood i feel

Dieter@be · 2009-04-20 11:22:39

ludovico wrote:

Dieter@be: Are you planning to write that program? I am no programmer, but
I would love to see that program realized, it sounds fantastic!

Maybe, if I find the time

Barrucadu · 2009-04-20 13:41:53

Dieter@be wrote:

I really like the unix philosophy, and I like simple storage backends which you can easily version/merge/synchronize. (eg plaintext)
I was thinking about a browser we could write with these principles. Let's call it foo for now
- each instance of foo renders 1 page (eg it's a small wrapper around webkit), no tabbing, tab previews, or speed dial things. we have window managers for that.
- 1 control application (lets call it fooctrl). use this to modify the behavior of a foo instance (change url, refresh). use xdotool to get the window with focus. eg fooctrl -win <id> -url <http://>. obviously the foo instances listen to basic signals too so you can use sigterm, sighup, sigusr1, sigusr2 ,... for common operations. (maybe refresh and back for usr1 and usr2)
- no keyboard shortcuts builtin.. use xbindkeys to bind keys to call fooctrl.
- no bookmark management builtin. make your own solution. for pulling a bookmark a plaintxt-based program using dmenu would work great here. combine with footcrl and xbindkeys. fooctrl should support an option to query the current page so you can script something to add to your bookmarks. use zenity or something to add tags.
- similarl story for history.
- no ad blocking built in. use the power of /etc/hosts. though fooctrl should support an option to list all images on a page, so you can easily pick the links to ads to add them to your /etc/hosts. (dmenu can again be great here to automate this)
- no download manager. allow user to pick wget/curl/a custom script/...
- etc...

Sounds a brilliant idea, unfortunately Python is the only language I know well, and for a browser designed to be small and use as little unnecessary resources as possible, that's not really ideal. I looked into pywebkitgtk, though the demo provided doesn't work fully and used 50% CPU on my laptop, making me think that another language would be required (though it may just not have been programmed very well).

If only I knew C++ or something, I could make a start on this…

Last edited by Barrucadu (2009-04-20 13:43:22)

Aprz · 2009-04-20 14:00:36

Heh, ever since I really got into Linux, I've always been a little anxious about making my own web browser even if it is as simple was links and not nearly as good. I've already done the same for a shell which some of you guys might of saw in the programming section of the forums here, still working on that by the way, now using git and chatting with people on that instead of posting it in the forums since it didn't get enough attention there. Maybe one day, I'll try to pull that off.

Anywho, I think all of you missed the biggest and most important feature that all web browsers are missing... to make coffee. If a web browser doesn't make coffee, how can it even be considered a web browser? <_<

thisllub · 2009-04-20 17:08:43

leo2501 wrote:

thisllub wrote:
It seems to me that the biggest flaw of Firefox is extreme memory usage when large numbers of pages are opened.
This could reasonably be solved if it implemented disk based persistence for non active tabs.
It's SLOW... lol, in my intel atom 1.6 it lags several seconds to switch between tabs or open a page in a new tab, even gmail is painfully slow... with links -g for example, browsing the internet is blazingly fast, something is slowing it down under the hood i feel

I am starting to suspect a problem with the Linux build.
I am typing this from my laptop which is OpenSolaris, wireless and with more than 20 tabs open I am seeing no performance problems. I am also running an Arch VM.
On my desktop I regularly have problems with long pauses. It is Arch with a faster processor than the laptop.

I definitely need to run some tests.

Barrucadu · 2009-04-21 16:58:53

Idea: Each instance of 'foo' could create a FIFO in /tmp, and the behaviour can be controlled by echoing things to it. For example:

echo "http://www.google.com" > /tmp/foo/1

could cause instance 1 of foo to load google, and

echo "refresh" > /tmp/foo/2

could cause instance 2 of foo to refresh the page.

Procyon · 2009-04-21 17:08:01

I wouldn't use a FIFO because of the delay, but a socket.

Barrucadu · 2009-04-21 17:11:19

Ah, I don't really know much about interprocess communication. If Dieter@be doesn't make this some time, this will probably be an interesting enough project to motivate me to learn C++

Dieter@be · 2009-04-21 18:36:51

Barrucadu wrote:

Ah, I don't really know much about interprocess communication. If Dieter@be doesn't make this some time, this will probably be an interesting enough project to motivate me to learn C++

well I have been playing a bit with gtk and webkit. you may see some code online soon.. or maybe not

oh and btw: scratch the "no keyboard shorcuts" point. that was too extremistic. I think I'll allow to configure basic keyboard shortcuts through an ini config file, but still i want to expose a lot of functionality through the fooctrl thing

Last edited by Dieter@be (2009-04-21 18:38:47)

Procyon · 2009-04-21 18:58:41

Do you plan to make the control window a permanent window, or something more like mpc?

manoa · 2009-04-21 19:34:38

leo2501 wrote:

thisllub wrote:
It seems to me that the biggest flaw of Firefox is extreme memory usage when large numbers of pages are opened.
This could reasonably be solved if it implemented disk based persistence for non active tabs.
It's SLOW... lol, in my intel atom 1.6 it lags several seconds to switch between tabs or open a page in a new tab, even gmail is painfully slow... with links -g for example, browsing the internet is blazingly fast, something is slowing it down under the hood i feel

it's because of the version, if version 1.5.0.12 compiled without errors I would use it, firefox 1.5.0.12 is the best browser, with a few modifications it's almost as fast as Dillo, I keep wondering why this version of firefox is not available in Arch, newer versions of firefox are much slower than 1.5 than 1.0 versions are slower than 1.5

firefox 1.5 all the way, only until Dillo gets it going right.

Barrucadu · 2009-04-21 19:34:43

Dieter@be wrote:

well I have been playing a bit with gtk and webkit. you may see some code online soon.. or maybe not

Well, I've been motivated anyway. Too late now.

Dieter@be · 2009-04-21 19:41:41

Procyon wrote:

Do you plan to make the control window a permanent window, or something more like mpc?

fooctrl (or uzblctrl as it's called now) will be comparable to mpc. it's not graphical.
Oh yeah http://github.com/Dieterbe/uzbl/ ..
current state? git clone, make and "./uzbl http://archlinux.org". see for yourself.

scio · 2009-04-21 19:49:36

Bleh, GTK. Don't make me rewrite it in Qt.

Dieter@be · 2009-04-21 19:51:53

scio wrote:

Bleh, GTK. Don't make me rewrite it in Qt.

Wait a minute. I don't need gtk or qt. just webkit should do. I'll see if I can get rid of it

Barrucadu · 2009-04-21 19:59:02

Using it now

I think I'll give up working on my own and just hack yours if i get the urge

manoa · 2009-04-21 20:08:12

GTK is great, as long as it's not version 2

Dieter@be · 2009-04-22 19:33:11

Barrucadu wrote:

Idea: Each instance of 'foo' could create a FIFO in /tmp, and the behaviour can be controlled by echoing things to it. For example:
echo "http://www.google.com" > /tmp/foo/1
could cause instance 1 of foo to load google, and
echo "refresh" > /tmp/foo/2
could cause instance 2 of foo to refresh the page.

Procyon wrote:

I wouldn't use a FIFO because of the delay, but a socket.

Great idea guys. I'll look up the (dis)advantages to both approaches and implement something like it. This makes a sweet interface, which probably means we don't need a separate control application.

To avoid disturbing this thread too much, I created a new thread for uzbl. http://bbs.archlinux.org/viewtopic.php?id=70700

Last edited by Dieter@be (2009-04-22 20:00:09)

finferflu · 2009-04-24 10:22:05

While I loved the new uzbl, I think the basic Unix philosophy could be expanded even more by separating the form from the contents. What I have been thinking about these days is a "semantic" web browser, which is not a simple "page visualiser", but something able to organise the content of pages the way you want, and also to pipe the content you want wherever you like.
I think this wouldn't be so hard with simple HTML, but it starts to be a hassle with the other technologies, like Flash etc. Obviously I don't know much about web browser and language parsing, but I'm thinking about digging deeper into it, since these days seems like a great deal of applications has moved over the web.

But for the time being, I'm pretty enthusiastic about uzbl

catwell · 2009-04-24 11:14:12

While we're at it, I dream of a browser that could natively handle small markup languages like Markdown (without converting it to HTML first). That shouldn't be too difficult to make and would work great with this "content-based" idea.

That would be the first step to a simpler (as in KISS) web. Most of us hate XML-based syntax for most things (configuration files...), why do we find it so great for the web? To me, human-readable source code for web pages would make a lot of sense.

Last edited by catwell (2009-04-24 11:14:55)

Dieter@be · 2009-04-24 14:19:58

finferflu wrote:

While I loved the new uzbl, I think the basic Unix philosophy could be expanded even more by separating the form from the contents. What I have been thinking about these days is a "semantic" web browser, which is not a simple "page visualiser", but something able to organise the content of pages the way you want, and also to pipe the content you want wherever you like.
I think this wouldn't be so hard with simple HTML, but it starts to be a hassle with the other technologies, like Flash etc. Obviously I don't know much about web browser and language parsing, but I'm thinking about digging deeper into it, since these days seems like a great deal of applications has moved over the web.
But for the time being, I'm pretty enthusiastic about uzbl

that's quite interesting indeed. But what are the use cases? Something like embedding a piece of a webpage in a different document (which could be pdf/openoffice/..) ? stuff like that?

Zariel · 2009-04-24 17:23:45

Why not use a vimerator style status bar?

timetrap · 2009-04-27 22:08:06

I am so impressed by what you have going here. When I first saw this thread, I thought, "Oh great, another discussion going nowhere." But damn, your uzbl browser is looking good. Simple, but I am really looking forward to using it. I don't know too much C, but might I make a suggestion.

Make a plugin framework in python.

That would be awesome. I'll start using uzbl from now on for my regular web browsing, and switching to firefox when I need flash, etc.

Procyon · 2009-04-27 22:11:31

uzbl can do flash just fine.

Arch Linux

#51 2009-04-20 02:34:39

Re: Arch Philosophy/Structure Applied to a Browser

#52 2009-04-20 02:44:49

Re: Arch Philosophy/Structure Applied to a Browser

#53 2009-04-20 11:22:39

Re: Arch Philosophy/Structure Applied to a Browser

#54 2009-04-20 13:41:53

Re: Arch Philosophy/Structure Applied to a Browser

#55 2009-04-20 14:00:36

Re: Arch Philosophy/Structure Applied to a Browser

#56 2009-04-20 17:08:43

Re: Arch Philosophy/Structure Applied to a Browser

#57 2009-04-21 16:58:53

Re: Arch Philosophy/Structure Applied to a Browser

#58 2009-04-21 17:08:01

Re: Arch Philosophy/Structure Applied to a Browser

#59 2009-04-21 17:11:19

Re: Arch Philosophy/Structure Applied to a Browser

#60 2009-04-21 18:36:51

Re: Arch Philosophy/Structure Applied to a Browser

#61 2009-04-21 18:58:41

Re: Arch Philosophy/Structure Applied to a Browser

#62 2009-04-21 19:34:38

Re: Arch Philosophy/Structure Applied to a Browser

#63 2009-04-21 19:34:43

Re: Arch Philosophy/Structure Applied to a Browser

#64 2009-04-21 19:41:41

Re: Arch Philosophy/Structure Applied to a Browser

#65 2009-04-21 19:49:36

Re: Arch Philosophy/Structure Applied to a Browser

#66 2009-04-21 19:51:53

Re: Arch Philosophy/Structure Applied to a Browser

#67 2009-04-21 19:59:02

Re: Arch Philosophy/Structure Applied to a Browser

#68 2009-04-21 20:08:12

Re: Arch Philosophy/Structure Applied to a Browser

#69 2009-04-22 19:33:11

Re: Arch Philosophy/Structure Applied to a Browser

#70 2009-04-24 10:22:05

Re: Arch Philosophy/Structure Applied to a Browser

#71 2009-04-24 11:14:12

Re: Arch Philosophy/Structure Applied to a Browser

#72 2009-04-24 14:19:58

Re: Arch Philosophy/Structure Applied to a Browser

#73 2009-04-24 17:23:45

Re: Arch Philosophy/Structure Applied to a Browser

#74 2009-04-27 22:08:06

Re: Arch Philosophy/Structure Applied to a Browser

#75 2009-04-27 22:11:31

Re: Arch Philosophy/Structure Applied to a Browser

Board footer