updates Meet the browser: Firefox Next

[primexx]

Probably, multiprocess is a pretty big change.

 

well... crap. lots of breakage incoming then.

[The_Decryptor Veteran]

The APIs that allow for multiprocess extensions have been around since Firefox 4, there's already a bunch of extensions that work fine since they used these "newer" APIs.

Extension breakage is to be expected, but it'll be old unsupported extensions (Which have their own issues), or badly written ones (Again, issues)

[SarK0Y]

the more addons you enable, the more crappy ff runs. multi-processing seems a proper way to make browser more retarded because you get additional latencies for inter-page websites.

 

Memory usage on the other hand is similar to single-process Firefox. The overhead is just 10 Megabytes in comparison, and the developers have stated that they may be able to reduce it further.

That's definitely good news considering that Firefox's excellent memory usage won't inflate because of this feature.

 
 
http://www.ghacks.net/2013/12/05/multi-process-firefox-go/

this part looks kinda suspicious: to reduce memory overhead at given case is possible only thanks to a common core to share basic functionality for opened tabs. in short, that scheme ain't pure multi-processing. so, i doubt there will be greater stability & security.

[ViperAFK]

yay, 27 beta fixed the annoying graphics corruption in the addressbar I was getting in 26  :)

[SarK0Y]

Viper, graphic errors are such a bliss, they can be pinpointed quite soon :)

[The_Decryptor Veteran]

...

this part looks kinda suspicious: to reduce memory overhead at given case is possible only thanks to a common core to share basic functionality for opened tabs. in short, that scheme ain't pure multi-processing. so, i doubt there will be greater stability & security.

It's called shared memory, 3 different instances of Gecko shouldn't take 3x as much memory, they can share all the executable code in memory (And anything else common between pages, 2 pages in 2 different Gecko processes referencing the same image should only store 1 copy of the image)

Edit: And the all cache/net access goes through the parent process (Sandboxing reasons, the child processes shouldn't have random internet/drive access), that reduces per-child memory use as well (A decoded image lives in the parent process, each child that uses it maps it through shared memory, etc.)

[mancode20]

Anyone have user style code for larger toolbar icons? They are too small! That's my only small complaint with Australis ( UX Nightly 29)

[Konstantine]

Any way to add colours to the icons(like "Addons", "Options", "Private Window" etc)? They are plain gray. :(

[The_Decryptor Veteran]

Yes, draw new ones (And for the post above, draw bigger ones)

[Konstantine]

Yes, draw new ones (And for the post above, draw bigger ones)

 

<snipped>

 

There should be a stylish code or something like that.

[The_Decryptor Veteran]

Sure you could replace the icons via a user style, but the fact is those icons are just monochrome, you'll have to draw your own colourful ones if you want colours (Or edit the existing ones). They were originally going to use SVG paths with a filter to do all the fancy gradient/dropshadow stuff, but that ended up way too slow (Maybe they should re-investigate that now that they have a rendered SVG cache)

Oh, and for the larger toolbar icons post. Right after posting I realised that there are already larger scale icons for HiDPI monitors, switching to those and specifying a larger size should work fine.

[SarK0Y]

It's called shared memory, 3 different instances of Gecko shouldn't take 3x as much memory, they can share all the executable code in memory (And anything else common between pages, 2 pages in 2 different Gecko processes referencing the same image should only store 1 copy of the image)

yes, Decryptor, shared memory o/& common-core process have been very way to reduce memory overhead. but such approach doesn't improve stability & security + it pounds the speed.

Edit: And the all cache/net access goes through the parent process (Sandboxing reasons, the child processes shouldn't have random internet/drive access), that reduces per-child memory use as well (A decoded image lives in the parent process, each child that uses it maps it through shared memory, etc.)

 

 

script blockage is better method to avoid viruses from untrusted resources, sandboxing ruins speed so badly much. in short, i vote for single-process scheme :)

[The_Decryptor Veteran]

Yeah it does, the rendering processes are separate, they can't take down the parent process and it lets JavaScript run concurrently.

Why would sandboxing the rendering process effect speed?

[SarK0Y]

Yeah it does, the rendering processes are separate, they can't take down the parent process and it lets JavaScript run concurrently.

Why would sandboxing the rendering process effect speed?

sandboxing traces code to prevent malware activities, so it needs a huge bulk of extra ticks on CPU & extra data moving as well. + needless to mention, JS code ain't very fast even w/o any sandboxes. + sandbox cannot guard against all kinds of malicious code & can crash process as well.

[The_Decryptor Veteran]

No, the sandbox they're using is OS enforced security boundaries and patching the code to prevent access to system commands.

It doesn't "trace" the code, it just stops it from being able to do things.

[SarK0Y]

No, the sandbox they're using is OS enforced security boundaries and patching the code to prevent access to system commands.

It doesn't "trace" the code, it just stops it from being able to do things.

sandbox runs in the simplified form like:

 

1. recognize action.

1.1. if action is pure threat, prevent its execution.

1.2. if action is possible threat, inform user about.

1.3. if No threat detected, pass through.

======================================

ideally, considers that sandbox cannot be visible by testified code. but(!), in fact, no sandbox can emulate environment for 100%, so code could conceal its dark sides + in some cases, it's possible to crash sandboxing from testified code.

[The_Decryptor Veteran]

I think you're thinking more of an AV trying to discover whether an executable is bad or not, because sandboxes don't work like that.

When you sandbox a process on Windows (As IE, Chrome and soon Firefox will do), what they're doing is making the client process run at a lower permissions level where it doesn't have user level hard drive access, no "emulation" or "tracing" going on.

Edit: Apps run as a user don't have the same amount of permissions as an Administrator, etc. It's the same system.

[SarK0Y]

Decryptor, thanks for pinpointing that difference :) i'm too stuck in av sandboxes. yes, low permissions are useful thing, but, for many cases, script blocking is much more preferable way: sandboxing, aforementioned by you, corrupts script logic for many cases as much as simple blocking, but blockage is more secure approach.

[wasd-]

Im having a problem since yesterday, if i have Greasemonkey enabled(Even with no scripts installed)Firefox doesn't start, i need to finish the process through task manager. I tried to uninstall it and try again but nothing. Any ideas how can i solve it without creating a new profile or resetting all?

[wasd-]

I fixed it. I deleted every entry in about:config with contained greasemonkey and now works fine

[kilara1988]

I had exactly the same problem as wasd- and fixed it by entering in safe mode and deleting any greasemonkey preffs.

[trag3dy]

I fixed it. I deleted every entry in about:config with contained greasemonkey and now works fine

 

 

I had exactly the same problem as wasd- and fixed it by entering in safe mode and deleting any greasemonkey preffs.

 

For future reference there is an addon that will delete any preferences for addons, it's called eCleaner and you can get it here..

 

https://addons.mozilla.org/en-US/firefox/addon/ecleaner/

 

Simplifies the entire process.

[Konstantine]

Bump guys. Ux hasn't been updated for over 10 days. :O

[The_Decryptor Veteran]

Why would it? Australis has landed on the trunk, they're not going to be doing theme work on temporary branches now.

[Konstantine]

There are things that must be fixed. Under-the-hood for example. Does that mean we should install Nightly? -.-