Mozilla Firefox 21 Hits The Web With New Features

[smitty3268]

Indeed, I haven't seen Firefox crash on Linux for at least a few years now.

About the update, I wonder if there were any improvements in PDF rendering speed. It's quite annoying as it is.

JS is going through some pretty major performance boosts right now, as they focus on improving the new ionmonkey compiler.

On that note, this bug has an interesting patch, although it'll be a while before it reaches stable releases. (50% pdfjs boost in the octane test)

https://bugzilla.mozilla.org/show_bug.cgi?id=872020

[Zan Lynx]

That is another issue with browser flame wars. The browsers seem to behave wildly differently in different hardware/software environments, leading to meaningless debates with anecdotal evidence and everyone believing that they must be right because they can see it with their own eyes.

Yes.

Personally, I think the biggest performance difference in Firefox on Linux is caused by differences in disk and filesystems. Firefox uses an embedded SQL database and occasionally requires a sync to disk to ensure data consistency. If you have a very slow disk (like a laptop drive that spun down for powersaving) or if your filesystem is EXT3 with the journal in ordered mode you can hit multiple seconds of disk wait.

[Greijoan]

Yes.

Personally, I think the biggest performance difference in Firefox on Linux is caused by differences in disk and filesystems. Firefox uses an embedded SQL database and occasionally requires a sync to disk to ensure data consistency. If you have a very slow disk (like a laptop drive that spun down for powersaving) or if your filesystem is EXT3 with the journal in ordered mode you can hit multiple seconds of disk wait.

From a very strange reason unknown to me, it seems that a lot of people think that people who rant about FF's speed have a problem with their computers, or can't differentiate between laggy GUI and changing disk power consumption states. I'm using multiple boxes with multiple operating systems. From old Pentium 4, through Q9300 and i5 M560, Phenom II 1090T, ArchLinux, Fedora, OpenSUSE, Windows XP, 7, 8, desktops & laptops, all setups clearly show me that FF experience is more laggy than Chromium or Opera. While FF can get more points in javascript benchmarks, and that's cool, but that doesn't change the fact that sometimes I can't scroll the page down because the other website I just opened in the background. Also, I'm wondering, did you actually run any profiling to see that most latency is due to the things you've proposed (power management + ext3)?

So I don't know. Maybe you're just more tolerant for slowdowns and you don't see them as irritating as me.

[smitty3268]

From a very strange reason unknown to me, it seems that a lot of people think that people who rant about FF's speed have a problem with their computers, or can't differentiate between laggy GUI and changing disk power consumption states. I'm using multiple boxes with multiple operating systems. From old Pentium 4, through Q9300 and i5 M560, Phenom II 1090T, ArchLinux, Fedora, OpenSUSE, Windows XP, 7, 8, desktops & laptops, all setups clearly show me that FF experience is more laggy than Chromium or Opera. While FF can get more points in javascript benchmarks, and that's cool, but that doesn't change the fact that sometimes I can't scroll the page down because the other website I just opened in the background. Also, I'm wondering, did you actually run any profiling to see that most latency is due to the things you've proposed (power management + ext3)?

So I don't know. Maybe you're just more tolerant for slowdowns and you don't see them as irritating as me.

Recent firefox versions have made a point of moving disk accesses off the main GUI thread and into async background tasks. Precisely so that when you open up new tabs, hitting the disk cache, history db, etc. doesn't lag the GUI.

They've made a point of targeting these kinds of issues since around 16 or 17, and they're starting to really make a dent in them now.

Also, yes, people have profiled things and ext3 had some known issues with slow fsyncing and the way mysql was being used. Obviously it's not the only issue, but it did tend to be one of the major issues Firefox had on linux. It was less of an issue on other filesystems.

[randomizer]

Also it seems mozilla is improving a lot firefox on the light of upcoming FireFox OS, they added JIT javascript, improved memory consumption, startup times, rendering performance, etc...

They actually added JIT compiling for JS back in FF 3.5, but yes, many of the other things have been focused on much more in recent times.

[uid313]

They actually added JIT compiling for JS back in FF 3.5, but yes, many of the other things have been focused on much more in recent times.

Yeah, but the JavaScript engine have been improved and rewritten a couple of times since.

Previously TraceMonkey then J?gerMonkey.
Now currently IonMonkey.
Upcoming is OdinMonkey.

[Serge]

I actually find the three-state DNT to be the most interesting change. It's not impressive or exciting from a technical perspective, but I interpret it as a strategic victory for privacy advocates in their tug-of-war with advertisers, and that makes it a very interesting topic for discussion.

The "three states" are to send no header, to send a DNT header, and to send a "do track me" header. Up to now, DNT headers were only sent on an opt-in basis. "Nothing" was the default, implicit state for those who had not exercised opt-in for DNT. Advertisers' public position has been "If the user does not send a DNT header, that means the users wants us to track them." Mozilla creating explicit options for both "I want to be tracked" and "I don't want to make a choice either way" is really going to turn things upside down for advertisers who try to claim that no DNT header implies a desire to be tracked.

Does anyone else have comments on the DNT changes?

Reference:

[aironeous]

For this release the score at html5test.com went up by 5. The new score is 399 the same as chrome 18
Score Bonus
Chrome 25 ? 463 13
Maxthon 3.4.5 ? 457 15
Chrome 23 ? 448 13
Chrome 24 ? 448 13
Chrome 22 ? 434 13
Chrome 21 ? 431 13
Chrome 20 ? 418 13
Chrome 18 ? 399 13