Slashdot Mirror
Linux: Browser Wars
Anderson Silva writes "LinuxToday has an article doing a pretty basic comparison on some of the major linux browsers. Although a nice article, and with a fair result, I still think Opera is the best browser available for Linux." I prefer knoqueror, although recent builds seem to have random hangs on images.google.com.
We both like 'knoquers'.
---- http://www.opedog.com/
I guess slashcode still doesn't include a spellchecker.
Somewhere in the heavens... they are waiting.
From the article: Opera is slick, but it's page rendering is nothing short of horrendous. Galeon performed well in all tests, and, aesthetics aside, it's a good choice.
I haven't noticed this myself...In my experience Opera has (almost always) been very fast in rendering HTML for viewing. Its only problem is that it waits for images to load before it displays anything past the image tag in question. Perhaps this was why it took so long to load the page in the test.
This proves once again, that there still isn't a good browser for Linux. So we have to decide on which one is less crappy, and not which one is better.
I don't understand why this is so. It sickens me that browsing on windows with IE is more stable then anything on the linux platform. Its just not right.
The author says Opera is clean and simple. In my eyes, Opera is horrible. It's default screen is covered with 500 different widgets. When you load a page, they all start whizzing and moving around. It's very distracting. Opera doesn't look at home on GNOME nor KDE, which just adds to its problems. Opera, with its adverts and grotesque widgets, is a visual insult.
% echo "GET / HTTP/1.0" | nc slashdot.org 80 | head | grep '^X-' | grep -v '^X-Powered'
X-Bender: Oh, so, just 'cause a robot wants to kill humans that makes him a radical?
X-Bender: OK, but I don't want anyone thinking we're robosexuals.
X-Fry: Nowadays people aren't interested in art that's not tattooed on fat guys.
X-Bender: Honey, I wouldn't talk about taste if I was wearing a lime green tank top.
X-Bender: Bite my shiny, metal ass!
Slashdot uses a lot of tables on the pages which can take a while to render, so what better test for a browser. Obviously this is a pretty extreme test as most pages are nowhere near as big.
Opera: 127 seconds
Konqueror: 57 seconds
Mozilla: 71 seconds
Galeon: 64 seconds
Skipstone: 57 seconds (Note: Browser crashed on first attempt.)
Netscape: 34 seconds
Winner: Netscape Navigator
These load times are absurd. Is this guy connected to the internet via a 300-baud phone-coupler attached to a telephone line spliced together with paper clips? I'm on a cablemodem, and it takes less than two-seconds to fully load slashdot. I think it took about 9 or 10 when I was on a dialup. Anyone else think these figures look a little inflated?
when salmon are outlawed, only outlaws will have salmon
The Mozilla version shipping with Mandrake 8.0 is 0.8.7. While stability is pretty much unchanged since then, Mozilla has gotten noticably faster during the 0.9.x cycle. 0.9.1 is usable on a 350 Mhz Pentium II... sort of. 0.9.3, while still being slower than Navigator 4.77, isn't bad at all. It's finally fast enough that I can use it as my normal, day-to-day browser (I was using Nav 4.77, because while it was unstable as hell, at least I didn't have to wait 20 seconds for a page to load).
I imagine that simliar situations are true for at least one or two of the other browsers compared. Development on Mozilla, especially, is happening very fast and comparing something current 6 months ago is not, IMHO particularly meaningful.
A German magazine did a similar thing a while ago, only they included MSIE. It won hands on in every discipline from speed to adherence to standards.
A pity that it wasn't at least mentioned.
Censorship on Slashdot
I'm using Galeon to read this right now 0.11.0, and while it's a really nice, clean interface, it does have some problems.
. htm
:)
http://ska.about.com/library/cannabis/blccrolling
This is one of them. One of the two pop ups on this page crash it EVERY time. Without fail. I warn you, do not visit this in Galeon (unless there is some way of turning pop-ups off, which is entirely possible, I've never really delved too deep into it's guts.
But I like it MUCH better than Mozilla and Netscape. It just seems cleaner to me.
And for those of you visiting that web site in non Galeon browers, I did eventually figure out how to roll a joint without its help
Verloc
From the article: Konqueror: Clean, simple and boring. Perfectly functional, with the bare minimum of fuss. The spinning KDE logo in the corner looks very nice, but of course adds nothing to your browsing experience.
Konqueror, boring? Gimme a break. It's completely themable and it doesn't even need its own themes like Mozilla, you can use general KDE themes. And it works wonderfully as a file manager (and network browser and PDF and manpage viewer), with smooth icon previews of HTML, ps, pdf, images and text files. You can split the view in however many sub-windows you want, you can even have a shell prompt as a subwindow. It has a full screen mode. Right now, I'm browsing with KDE and Konqueror in "Aqua" theme and it looks, well, let's just say you have to buy an Apple if you want something to look cooler than that.
And what's up with testing on a ridiculously outdated machine? P166, no MMX, 32 MB RAM? You've gotta be kidding me. If I wanted a browser that worked fast on this configuration, I'd have stuck with Netscape 3.0...
I'm sorry, but aren't the versions of the browsers used somehow important to the story? Was it Mozilla 0.9, or 0.9.2? Netscape 4.08 or Netscape 4.7, or Netscape 6? Hard to tell what these tests mean, especially if not the latest versions of each browser are being used.
icqqm [ICQ:11952102]
why oh why where text only browsers not included ?
is this because their user base is small ?
I personally use it but I find that alot of people dont
because I find lynx the fall back GOD the page doent render in netscape or some fool has FSCK the HTML I just use lynx and away I go
really how much information (I am intrested in )is presented in pictures on the web
not much I am sure
lynx is my fallback king (-;
I use it when I telnet into places to check they can see stuff plus all I need is a telnet app which I can obtain for most OS's
what do you relie on to ALWAYS give you the web ?
(me its a telnet client and lynx)
regards
john jones
Hey, at least I didn't post a lame joke about the obvious misspelling. Get a life, people, willya?
Stating on Slashdot that I like cheese since 1997.
*CmdrTaco loads images.google.com*
*CmdrTaco types in "tux the penguin nude"*
*Penguin loads up in goatse position*
CmdrTaco: hmmmm....
*CmdrTaco types in "RMS nude"*
*Google locks...*
CmdrTaco: ^$%$#@!
Execute? [Y/N] _
One of the most glaring is that it won't render table cells with no content, so you have to put a non-breaking space in every empty cell. It also screws up table widths.... I could go on and on... ask anyone who works on web application development, they will tell you, Netscape sucks.
If it looks good on your machine in Netscape, it's only because someone slaved away to make it that way.
Being a programmer does not make you intelligent.
The inverse is also true.
Sorry, the amount of memory used by XFree86 isn't really all that much. What you're seeing when you see huge memory usage for X in top is because the X process has memory mapped your video card's graphics RAM into its memory space, several times over.
On my 32 meg GeForce2MX card, top shows X taking up 135megs of RAM. On a friend's system with an old school 2 meg VRAM card, X is only shown taking up 4-5 megs of RAM.
X is actually pretty damn memory efficient. Remember it was originally created when a workstation might have had one megabyte of memory, total. If you have a lot of windows open at high color depth, there will be some real RAM taken up to store those bitmaps, depending on whether you have 'save unders' enabled, but that's a function of all of the programs you have running, more than of X's inefficiency, even if the memory is counted against the X server process and not the X programs themselves.
FWIW.
I still think that the browser tests covered here are rather meaningless on a 32 meg machine. These days, browsers will take up close to a full 32 megs of RAM on a UNIX system, especially with the 'cache in RAM' option of Mozilla and Netscape. These days, when you can get 512 megs of PC133 RAM for less than fifty bucks, it just doesn't make sense to worry about 32 megs here or there, anymore.
- jon
Ganymede, a GPL'ed metadirectory for UNIX
The Winner [for "The Look"]: Mozilla, hands down. It's terrific that someone decided to take the route away from the greys.
Oh goody. I was tired of all my applications looking the same and behaving the same. I love guessing which color means disabled for each different application. I like having my system wide colors that I've carefully chosen to minimize eye strain thrown out the window.
System wide colors and looks are feature. If you're sick of living in grey land, change it globally. Gnome supports this. KDE supports this. Windows supports this.
Mozilla is a great browser, but their decision to roll their own user interface was a mistake. Fortunately Mozilla is modular, and as the core engine stabilizes I plan on moving to a more system friendly browser using that engine. Probably Galeon or Skipstone.
Search 2010 Gen Con events
Unfortunately this outdated-browser analysis is worthless, for several reasons - all of which boil down to the small amout of actual testing done. A faster, more representative machine would have been useful as well. And I don't need an analysis of browsers way back when on a machine from way back when and then some :-).
:-(
But should you doubt me:
First off, I think the one most deciding factor in the choice of a browser if how well it displays pages - whether corrupt, IE5.5 optimized, javascript enabled, CSS2.0 or ancient, my browser first and foremost needs to WORK. This isn't even touched upon here! The stability of the browser, in my opinion a part of usability, needs to be tested.
A browser doesn't need to be all that fast either just "fast enough". And, not only is "fast enough" a subjective measure, it includes things such as responsiveness while loading, total page loading time, time to create a new window, time to "scetch" a first outline onscreen and more. Many pages can be very usable with only 10% loaded. By the time you're done reading the first paragraph the rest can be loaded. In addition, speed will vary depending on processor speed and type, memory availability, and network bandwidth. A fast browser which gains speed with bad incremental display could be worse than a slower version in which you can start reading immediately. Furthermore, the internet extends beyond slashdot... some HTML elements may render in varying speed depending on the browser used.
Speed is a hard thing to measure. This analysis isn't nearly complete enough to be at all useful.
Startup time is effected by things such as program size (if too much else is loaded, a 32meg machine might well be swapping skewing the image drastically), speed ratio between hard drive and processor, and VERY importantly, dependance on shared libraries. Konqueror for instance might seem much faster when running KDE already... and the same goes for the other browsers too though I don't immediately know which libraries they use. Notice how fast those "second instances" pop up...
Finally, this is a pretty lame attempt to harvest slashdot links by using a slashdot page in a linux browser test...
It is great to see that he used a fairly low end system to do his tests. There are so many Pentium 75 - 200 systems around that are still perfectly useful if people would just think a little harder while programming, and it is nice that somebody is realistically putting one to the test with more "modern" software...
Posted from the wireless couch.
I had to do a Browser comparison with an application which needed hiding and displaying parts of a web page: depending on what you clicked some different stuff appeared.
This application uses a lot of features a browser can handle: stylesheets (and the nasty "display" attribute), JavaScript, tables, forms and XML.
I tried the following browsers (under Windows, since the people who will use it mainly have Windows):
Netscape 4.x
Netscape 6.1
Internet Explorer 5.x
Opera 5.12
Amaya 5.1
Mozilla 0.9.3
Here are the results:
- IE kicked ass in everything, and even displayed the XML stuff right.
- NS 6.1 kicked ass too, but 6 or 7 times slower. Prettier display, but hideously slow (and no XML, but we didn't care). Same thing for Mozilla (duh).
- NS 4.x sucked. Couldn't handle the "display: none" property properly. No XML.
- Opera faked kicking ass, but in fact had JavaScript problems... just wouldn't show anything whatever you clicked. No XML.
-Amaya didn't even fake. I guess it was a JavaScript problem because the display of the object was weird. But it faked some XML. displayed the source as plain text (ohh it's displaying something!! no, it's the source)
Conclusion: best results on Win: (sniff) IE. Followed by NS6.1 and Mozilla. Then comes Opera.
Gotta try some browsers under Mac and Linux now too, maybe.
E
a site that will crash a browser one day will work fine the next in my experience
No kidding. IE crashes on me multiple times daily, but I very rarely find a reproducible set of steps I can take to make it crash. Mozilla crashes on me occasionally, but I can almost always figure out what I need to do to reproduce the crash so I can file a bug.
That doesn't mean it's impossible to measure stability. It just means that being able to find reproducible crashes isn't the same thing as having a stable product. If you wanted to compare the stability of various browsers, you would have to get a group of users to try different browsers for their daily browsing while running your own crash reporting tool, but that's far from impossible to do.
Mozilla comes with a third-party program called Talkback that reports crashes to the developers. mozilla.org uses this data not only to find the most common crash bugs (by comparing the tops of the stack traces), but also to calculate theh "mean time between failure" to determine whether any given milestone (and maybe even nightly builds) is particularly stable. Internet Explorer 6.0 comes with a similar feature. (Both Mozilla and IE6 prompt the user before sending the crash report.)
The shareholder is always right.
Frankly, if you care, rerun these tests yourself; I don't think the figures quoted are representative.
I'm old enough to remember when discussions on Slashdot were well informed.
Comment removed based on user account deletion
There's an IE 5/WINE howto at:
http://www.hardcorelinux.com/wine-howto.htm
which shows you how to run IE 5 in Linux. Someone wrote me recently stating v5.5 doesn't work w/ the command-line parameters i used, but I know personally circa 5.0 does. It works decently too, rendering pages nearly as well as the Windows counterpart.
SEO Copywriter. Just Say ON
I really don't see the big deal in designing an app that works in all browsers.
For me, it is part of my job, and yes I bitch about it but I am required to make it "work". Sometimes that involves kludgey work arounds, sometimes it involves designing two different versions of the same site, and sometimes it means dropping a bell there and a whistle here. Hell, at our office we have a guy who specializes in Netscape quirks, and he is great to have when things like this come up.