Firefox 3 Beta 3 Officially Released

firefoxy writes "Mozilla has officially released Firefox 3 beta 3. This release includes new features, user interface enhancements, and theme improvements. Ars Technica has a review with screenshots. 'Firefox 3 is rapidly approaching completion and much of the work that remains to be done is primarily in the category of fit and finish. There will likely only be one more beta release after this one before Mozilla begins issuing final release candidates.'"

Extensions

Looks good. All we need now are for the extension developers to make their extensions Firefox 3.0 friendly.

Re:Extensions

Exactly! You would think there would be some 'legacy plugin support' for people to enable if they so desire. I don't know that all of my plugins are being actively developed, and I cannot stand this version of Firefox on OS X for much longer (the beta is much more stable, but no plugins work)

Re:Extensions

I get them to work by setting extensions.checkCompatibility to false.

A few still refuse to work, but most do.

Now, can someone tell me how to keep my bookmarks always sorted by name? The two extensions I know of that do this job ignore my "don't check compatibility" instructions and still refuse to show up in the menus.

Re:Extensions

Exactly! You would think there would be some 'legacy plugin support' for people to enable if they so desire.

There is. Install the Nightly Tester Tools [oxymoronical.com] plugin. It adds a "Make All Compatible" button in your Add-ons dialog that does pretty much just what it says.

Firefox 3

I've been using Firefox 3 (trunk builds) before Firefox 2 was an official release. I love it.

Whatever happened to:

> Issue one major release every year (Fx 3 in 2007, Fx 4 in 2008, etc.) since it helps drive upgrades and adoption

http://wiki.mozilla.org/Firefox3/Firefox_Requirements#Release_Roadmap [mozilla.org]

Now my dream is to see a QT brand of Firefox again, perhaps using QT 4's built-in Webkit. Unify Konqueror, Safari and Firefox on one rendering engine and work towards making that the best damned rendering engine out there. They spent nearly two years on the new Gecko rendering engine, and it still isn't as fast as Webkit/KHTML. Firefox has all the features I want for the most part. I'm not saying they should abandon GTK, but they support multiple widgets and toolkits. Someone please give me a QT 4 branch of Firefox and I've be very happy.

Add-on finder?

One of the most promising and impressive new features in beta 3 is an integrated add-on installer system that allows users to search for and install add-ons from addons.mozilla.org directly through the add-on manager user interface.

Brilliant! Must build from trunk again!

Hints

First of all, I would strongly recommend actually uninstalling (completely) and reinstalling Firefox if you want to use this beta. Some apparent conflicts between my extensions for Beta 2 and this install caused some of the weirdest, buggiest behavior I've ever seen in Firefox. Only by wiping my profile and starting from scratch was I able to get tabbed browsing to work correctly.

Secondly, if you're annoyed by the new theme, just switch to Small Icons. It looks fine, except for the slightly annoying "Home" button.

Speaking of the "Home" button, it's on the Bookmarks toolbar now, in case you were wondering. You can move it back where it belongs while in the Customize Toolbar dialog.

So far, I don't see a whole lot to write home about. The new theme is definitely ugly. On the other hand, the beta feels very stable and very, very fast.

The feature I really want: whole-page zoom

The reason I want FF3 is to get whole-page zoom.

http://arstechnica.com/journals/linux.ars/2007/07/27/firefox-3-gets-full-page-zoom [arstechnica.com]

I use a 110 dots per inch monitor. I hate, hate, hate all web pages that were laid out with WYSIWYG design tools, with fonts set to 7 pixels tall and columns also specified as a certain number of pixels wide.

I don't have eagle eyes and I don't like to sit close to my screen. So I have my personal CSS forcing fonts to a minimum size... which makes some pages ugly, and other pages unreadable (depends on how much the page designer hard-coded with pixel sizes). I'm also using the ImageZoom extension to scale up images... which means the scaled images cover up lots of text on many web pages, and fancy graphical navigation buttons often don't match up with their clickable regions.

And I have a 16:10 ratio monitor... which means that often I will read a web site and there will be a narrow strip of text in the center, and tons of wasted space to either side, again because some web designer hard-coded things with pixel counts.

I used to wish that web designers would make sites that can adapt to unusual screen sizes. Well, the WYSIWYG tools aren't going away, so now I just want to zoom my pages.

steveha

Try it without installing with Firefox Portable

Mozilla Firefox, Portable Edition 3 Beta 3 was released a few hours after the announcement. It's packaged with a launcher so it runs self-contained so you can use it from a flash drive, iPod, portable hard drive, etc. But it's also handy for trying out the current beta without affecting your local install. You can even run it from your desktop to try it out and then delete it.

It's available from the Mozilla Firefox, Portable Edition 3 Beta 3 homepage [portableapps.com].

Re:So...

http://www.squarefree.com/burningedge/releases/trunk-for-firefox-3.html [squarefree.com]

• Memory leaks
  • 333078 [mozilla.org] - XPCOM Cycle Collector [mozilla.org]. (Cycle collection [psu.edu] has similar goals to tracing garbage collection [wikipedia.org] but integrates better with reference counting. Turning on cycle collection fixed entire classes of leaks, both in Firefox and in extensions.)
  • 330128 [mozilla.org] - Calling cancel() on a timer doesn't drop reference to callback.
  • Many more: 331 bugs fixed on trunk with the "mlk" keyword [mozilla.org].
• Code size and memory use
  • 296818 [mozilla.org] - Don't hold onto decoded image data for so long.
  • 143046 [mozilla.org] - Reduce memory use for animated GIFs by storing frames other than the first at the original 8 bits.
  • Take a string constructor out of line. (From 345517 [mozilla.org].) (1% code size win.)
  • 332174 [mozilla.org] - Drop SOAP support. (2% code size win.)
  • 313309 [mozilla.org] - Provide table-driven QI mechanism.
  • 407459 [mozilla.org] - [Windows] Switch from default MSVC malloc to jemalloc for better memory allocation speed and lower fragmentation.
  • Many more: 100 bugs fixed on trunk with the "footprint" keyword [mozilla.org].

In Short

Fast, really fast.

Don't trust me. Try it.

Re:So...

High memory usage is different from memory leaks - every time you open a new tab it stores in ram some of the previous and next pages in ram. So if you do a lot of surfing on different tabs it very quickly goes up to 100MB in ram. You can disable that from the settings but you lose the ultra-quick back and forward capability.

Re:So...

High memory usage is different from memory leaks - every time you open a new tab it stores in ram some of the previous and next pages in ram. So if you do a lot of surfing on different tabs it very quickly goes up to 100MB in ram. You can disable that from the settings but you lose the ultra-quick back and forward capability.

Yes, that's true, but it isn't the entire issue.

I'm running Firefox 3 (the previous beta, not the latest), and I set max_total_viewers to 1, which should in theory do what you said. Yet I routinely see ~200MB used by Firefox, and on my 512MB machine I need to restart Firefox once a day or so, since a web browser taking up half of my RAM doesn't make for good responsiveness of everything else.

One issue might be memory fragmentation in Firefox, or so I've been told. Perhaps someone who understands this stuff can clarify.

Re:If you close the tabs, does it free RAM or Leak

My typical memory-burning web surfing session is to go to Google News or especially to Fark.com, open up about 100 tabs of potentially interesting news stories, and then go read them one at a time, closing each one after I've read it. It's one thing to have the browser use lots of memory while I've got all the tabs open - but when I've finished with them all, and just have the original page back, or even hit "Home" to get "about:blank", the browser typically *still* has over 100MB of RAM and is often burning 20-70% of CPU.

I've never had Firefox use that much CPU, but many of those tabs you closed are still cached in memory (along with each of their histories) so they'll reopen really fast if you Undo Closed Tab. Closing the tabs does not necessarily mean they're going away. Changing this option [mozillazine.org] in your about:config should keep that from happening (I think), but you'll also lose some of your session restore functionality. I have it on, and I've never had any of the problems you and a lot of other people have, but I hope this helps.

Re:Adding bookmarks

Not that box, apparently there was a notice that appeared when you clicked the star to let you know what the star did (that the page had been bookmarked)... however the notice was both added and removed in nightlies between b2 and b3 (guess it ended up being more annoying than helpful).

Re:Is it faster?

FF3 is loads faster than FF2. I find that most slowdowns in FF2 were caused by extensions, but FF3 loaded with extensions is just as fast as FF2 in safe-mode. Which is fast.

Re:Is it faster?

While you're asking for it to be "faster", other people are asking for a smaller memory footprint.. considering that most performance issues in a browser are related to caching, they can't please all the users all the time.

Re:Is it faster?

While you're asking for it to be "faster", other people are asking for a smaller memory footprint.. considering that most performance issues in a browser are related to caching, they can't please all the users all the time.

Actually, the perceived performance issues of Firefox mostly stem from the fact it's a single-threaded architecture running on a JavaScript+XML interpreter (XULRunner).

Extensions, which basically "patch" themselves into this single-threaded synchronous engine, often exacerbate the problem too.

All XUL applications seem to share this slow response / performance problem, other popular ones exhibiting the same issue being Joost, Miro, SunBird.

However this issue is so deeply ingrained in XULRunner, that I hear misguided excuses all the time, such as "it's about the RAM cache / CPU usage balance", which, oddly enough, no other major browser suffers from (I use all on a daily basis as a developer myself).

About when we'll see improvements: most likely starting with Firefox 4, which is to completely replace the current JS engine, SpiderMonkey, with the one in Flash 9 (codenamed Tamarin), which compiles to machine code before execution, instead of being interpreted from opcodes.

We'll hopefully see some threading too (one thread for the main UI and one per tab at least), although the lead Mozilla developers have some quite irrational fears of multi-threaded architectures.

Threads are harder than you believe

Actually, the perceived performance issues of Firefox mostly stem from the fact it's a single-threaded architecture running on a JavaScript+XML interpreter (XULRunner).

There is indeed only one thread handling the UI and DOM, but there are multiple threads. Network operations, file decoding and so on run in separate threads from the UI. MAking a multithreaded UI is quite hard; note that IE (at least 6, most likely 7 too) does that too, with the difference that you can have separate windows in different processes altogether; but then they can't talk to each other through JS.

The only time this architecture is really a problem ATM is when JS from a page sucks up CPU: it bogs down the whole UI.

Moving to a fully multithreaded architecture is a very hard problem, esp. for such a complicated application, with such complex interactions as a web browser. Every single little thing would have to be synchronised, with big deadlock risks at each turn.

The only possible approach is to divide work among threads such as there is minimal, well understood interactions between them. You can't for example just have one thread per window, because HTML+DOM+JS expect to be able to touch other windows from the same domain. You could divide processes by originating domain; that's what Apprunner does.
But then you have coordinate communication between the windows and the bookmarks, history and so on. Not too hard to do, but has to be weighed against the minimal gain.

Eventually, we will have to take advantage of many-cores CPU. That means that even DOM parsing will have to be multithreaded, for use on ultra low power 256 cores mobile cpus. Robert O'Callahan [mozillazine.org] is working on this. But what you have in this case is a number of related threads with a very limited scope, and precisely defined interactions.

You can read more on these issues at his blog:

Parallel Dom Access [mozillazine.org]
Night of the living threads [mozillazine.org]

Re:Firefox 3 Mac OS X UI

Except... the problem with themes which try to emulate the native look and feel of the platform is that it has to be all or nothing; getting even a minor detail wrong can throw off the whole theme. This is even worse on the Mac, where there are a lot of users who are much pickier than average about the look and feel of the UI -- it has to match the native interface, because if it doesn't they're going to notice. And in the provided screenshots, I can already spot ways that the "native" OS X theme doesn't cut it. For example, the screenshot which proudly shows off an Aqua-style select control and button next to a search box also shows those controls using the wrong font and with the text incorrectly placed. If they can't get those details right, they might as well not try to do a "native" theme at all.

Re:Firefox 3 Mac OS X UI

If they can't get those details right, they might as well not try to do a "native" theme at all.

List boxes have always been ugly in Firefox. I don't think the theme has any control over this. Buttons look pretty good in 3.0 beta 3, but there are some nasty rendering artifacts on in the tab labels.

I agree with you that the details can make or break the experience. I keep trying to use Emacs shortcuts (Ctrl-A, Ctrl-E, etc.) in this text area, but this isn't a native control.

From what I've seen in the last fifteen minutes, 3.0 beta 3 is a big improvement. I've been pretty frustrated with Safari's performance. I'm not a kung fu memory master, but I do know that top shows up to 400 MB RPRVT and close to 2 GB VSIZE after it has been open for a while, even with only one or two tabs open. Sometimes when I close a tab it hangs indefinitely with a beach ball, so I have to force quit. If Firefox can spare me that annoyance, I'll forgive a few UI quirks.

Re:YAY!

immediately load urls I click on them from the address bar, instead of waiting for me to hit return

No. That's a terrible idea, and would drive innumerable people (myself included) completely crazy. Text-entry fields shouldn't do anything when you click into them in order to edit. The return key is the proper way to actually cause an action to be taken on the entered text.

That's a user interface paradigm that's decades old now, and just because the bunch of monkeys coding IE think it's fun to throw it out the window doesn't mean it's a good idea. Microsoft has the anti-Midas touch for interfaces these days anyway (cf. Vista generally, that new Office abomination generally, drop-down menus that hide half their contents for no particular reason, etc.). Emulating them would be a terrible idea.

Re:YAY!

I think they should improve the AddressBar/Dropdown to incorporate things like being able to remove a URL from the list

You mean like hovering the cursor over a URL in the list and pressing the delete key? Works nicely in 2.x

Re:YAY!

Yes I do, but how many average users are going to know that? I'd be willing to bet that the Parent didnt even know that...besides, thats only one of many features that its lacking in comparison to Opera, Avant, etc. There isnt much coding invlived to add a context menu on right-click, or even a tooltip saying "Yadda Yada for Yada!"

Re:acid 2?

Yes, Firefox 3.0 passes Acid 2 [ajaxian.com].

I'm hoping that they bring forward Tamarin support in Firefox. Any chance of getting fast javascript before Firefox 4?