About:Snappy

[ferongr]

Image boards don't represent normal web browsing.

What is "normal web browsing"? How is it defined? By rankings from web analytic providers? By some other objective measurement? By arbitary definition?

[iwod]

Any fruits, or planned seeds coming yet?

[patrickjdempsey]

Any fruits, or planned seeds coming yet?

Taras registered arewesnappyyet.com (on the 18th) but it's just an empty GoDaddy shell right now.

[_Alexander]

Image boards don't represent normal web browsing. And some sites like that don't actually use *thumbnails* they use *squished* versions of the full-size images, which is a bad bad bad practice.

None of the ones I surf do that resizing when showing thumbnails - it creates unnecessary server strain.
It represents a *great* example of how bad Firefox can be...

More _common_ webpages have a lot more flash content which also triggers a variety of issues from common (dropped frames) to more common on low-end systems (slow redraw)...

ARGH WHY IS FIREFOX HOLDING ALL THESE IMAGES IN MEMORY IT'S SUCH A MEMORY HOG

Devastating for *chan imageboards and *booru image galleries (easily threads of 200+ ~200x~200 img thumbnails)
Up to half a minute to restore / switch tabs!

Now use a usercript (e.g. 4chan X) that preloads or inline-expands those hundreds of images. You can do the math.

... or not use Firefox as my primary browser, a simpler solution chosen by normal folk.

[ferongr]

... or not use Firefox as my primary browser, a simpler solution chosen by normal folk.

Read what I wrote again. Inline-expanding or preloading images results in exhaustion of address space on 32-bit systems, and either extreme paging or OOM.

[Omega X]

If he's not using Firefox anymore, then why is he still here?

[_Alexander]

If he's not using Firefox anymore, then why is he still here?

I am waiting for a good browser with proper hardware acceleration and speed greater than Opera 11.6beta.
For no now no one delivered,
Opera 12 - HW slows things down 20x
IE9 - Very fast, but bad GUI, bad adblock, no spellcheck
Chrome - gets it ass kicked by Opera 11.5 / 11.6 with HW enabled

... or not use Firefox as my primary browser, a simpler solution chosen by normal folk.

Read what I wrote again. Inline-expanding or preloading images results in exhaustion of address space on 32-bit systems, and either extreme paging or OOM.

Ah, so you are proposing a way to have Firefox destroy itself! I read that initially as a script which introduces loading images on demand to reduce memory usage

[ferongr]

... or not use Firefox as my primary browser, a simpler solution chosen by normal folk.

Read what I wrote again. Inline-expanding or preloading images results in exhaustion of address space on 32-bit systems, and either extreme paging or OOM.

Ah, so you are proposing a way to have Firefox destroy itself! I read that initially as a script which introduces loading images on demand to reduce memory usage

It's handy. My point is, Firefox needs better ways to decide when to drop images from memory, even on the active tab. The browser should not be crashable, no matter what a webpage tries to do. It's one of the reasons that impact responsiveness. Modern CPUs can decode images when needed pretty fast (progressive decoding just outside the viewport is a way to minimize image related memory problems), so there's no responsiveness problem with progressive decoding. Address space for the process, on the other hand, is limited. There's also the issue where images are not CC'ed when removed from the DOM, resulting in similar problems on AJAX slideshows (Picasa and the like).

It's not a matter of "fringe cases" (like the 500th most visited site in the US or 1000th in the world is a fringe case...) but rather a matter that affects a multitude of image-related pages.

[patrickjdempsey]

The browser should not be crashable, no matter what a webpage tries to do.

mmmmkay.... crashing and responsiveness are two different things.

[ferongr]

mmmmkay.... crashing and responsiveness are two different things.

Depends on the system's available memory. On 2GB, you don't crash, but page out everything else resulting in everything freezing. On more than that, you may crash if you have the free memory, but it's not certain. This is disregarding the fact that when the heap grows beyond a certain point, my experience is that everything starts slowing down (due to CC pauses).

The underlying issues are the same, and affect multiple ways of interacting with the browser, and its perceived responsiveness.

[Omega X]

Patrick is right, crashing and responsiveness are two different sides of the coin.

A crash is a crash, and it happens immediately. That's stability. Delays in response, that's responsiveness. No response i.e. permanent freeze is practically crashing, that's stability again. Firefox is pretty stable, but there are obvious response items that need addressing.

[_Alexander]

Do not reply to this email. You can add comments to this bug at
https://bugzilla.mozilla.org/show_bug.cgi?id=539356

--- Comment #34 from Randell Jesup [:jesup] <rjesup@jesup.org> 2011-12-07 20:11:16 PST ---
[snappy] != perf
This is a perf bug. So far as I know, this is not a user-response-time bug.

discuss?

[KWierso]

discuss?

Project Snappy doesn't have clearly defined qualifications?

[mikedl]

discuss?

Project Snappy doesn't have clearly defined qualifications?

In my estimation, that would make some sense as it winds up being a highly subjective metric.

[Zlip792]

Blog post related to Project Snappy:
http://blog.mozilla.com/tglek/2011/12/0 ... ct-snappy/