Firefox/Waterfox and Broadband users, READ THIS!!

[Topcat]

^
Wow, the most I think i've ever seen it use was like 700mb....and that was with a buttload of tabs going on multiple windows....

[Scenic]

I ran into the 2GB 32Bit limitation on a regular basis till I found Waterfox (basically a 64bit ffx)

Old screenshot..

edit: a bit hard to click on the attachment.. so I [img]'d it

[Topcat]

^
Waterfox doesn't support XP64....I tried it.

[Scenic]

I guess the only option would be firefox nightly builds then..(?!)

No idea if those work on XP64 though..

example (no installer)
ftp://ftp.mozilla.org/pub/firefox/ni...n64-x86_64.zip

[walking]

Hmm, Nice.

Pages load like greased lightning now.

I notice there is like a 250 millisec delay then BANG the entire page is there.

My dsl is 1.26 and .38 up and loading is waaaay faster now.

[faisal1046]

Thanks...

[RJARRRPCGP]

Originally Posted by walking

Hmm, Nice.

Pages load like greased lightning now.

I notice there is like a 250 millisec delay then BANG the entire page is there.

You can take out that stupid delay. Don't need any dial-up-like 250ms bull.

In fact, I could ping less than 250 ms when I had 56K.

[shovenose]

This still works I'm surpised they haven't fixed these changes into the new versions.
Even on a lousy iMac Core 2 Duo running Windows 7 it's much snappier browsing.

[DJduck]

This is black magic Thanks, its faster now (using wifi)

[TELVM]

Couple extra tips for even more speeding of Firefox or any other browser.


Move the browser cache onto a ramdisk

or

Put the browser cache directly in memory

[mariushm]

So much bad advice and so much reviving of old threads... it's annoying.

Seriously, it's stupid to keep the cache in ram, if it was normal it would have been enabled by default.

Caching to ram prevents Firefox from caching stuff between browser restarts therefore nullifying anything that was cached in the session you just finished. So next time you start Firefox, you download everything because the cache is empty.

ffs leave the caching to disk, and leave it to write as much as needed. You have lots of crap cached between sessions, javascript resources, icons, logos, css stylesheets, facebook and youtube stuff, a lot.
Firefox will reuse them after you restart and save you bandwidth and improve the speed of rendering.

When Firefox will save to cache something, it uses basic windows functions to create the file on disk and to write the contents to it.
That means Windows (the OS and the NTFS drivers) intercepts the content that's written to disk and the location and besides delaying the write to disk (disk caching etc), it also keeps this content in RAM until other applications demand memory and the OS is forced to make room.

You won't get hard disk hits unless Windows was forced to remove that file from memory to make room for other applications or certain time threshold is reached or the file was touched by another application on disk (invalidates cache).



12 days of uptime, 10 GB of unused memory currently used by Windows to keep files I've accessed (read or written to) in cache for fast access... the Firefox cache in large majority is there already, the drive won't get hit for anything except updating last access time and crap like that, and that's done in a quick burst.


Besides this normal caching of various scripts and files that are explicitly marked as "can be cached by the browser", Firefox already keeps in memory by default the resources of the tab you're using, and the previous state (so you can click Back and get instant result).. so really when you click on a website on a link, 90% of the time the stuff that's supposed to be read from cache is already in memory from the fact that the page was loaded already.
So the cache helps most if you start the browser the next day and visit a website again, and ram cache screws this up, cause it's lost when you close the browser.
It's STUPID.


---------

And I think I've already said in this thread... the original advice is not really all that great:

Set "network.http.pipelining" to "true"

Set "network.http.proxy.pipelining" to "true"

Set "network.http.pipelining.maxrequests" to some number like 30. This means it will make 30 requests at once.

The value of 30 is too high and it has the potential to slow down the pages. Yes, you get the page all at a time, but much slower.

A much saner value would be about 8-10.

===
Explanation :

A website page is made of several files: the html file/code , the stylesheet or stylesheets, javascript files, pictures, other files

There are some hard limits in the browsers, by default a browser will only make 2-4 parallel connections to a domain or IP, so normally several connections are made and files are retrieved one at a time through each connection.
But it takes some time to create the connection to the site and request a file (script, css, js, picture, whatever), depending on how far you are from the server where the website is hosted it can take hundreds of milliseconds to a second.
So normally, it takes a bit to create the connection, request the file, close the connection. Repeat.

Pipelining simply changes this by not closing the connection and connecting again. The browser loses some time connecting initially, then the moment a file is downloaded, instead of closing the connection it just requests another file.

===

The maxrequests, this value, doesn't mean it makes 30 requests at once, but rather how many files the browser will download through a connection it created with the website. If you put 30 there, it just means that the browser will keep using that connection as long as less than 30 files were downloaded through that connection.

But the browser restricts the number of separate connections to a small value, 2 to 4 connections usually, so you actually want files to be pipelined to save time and not disconnect and connect so much, but you also don't want those four connections to be constantly used, or blocked by some slow transfers for a long time. Some ISPs also throttle connections after some amount of data is sent, so with such provider, you don't want to download 30 pictures on a connection, because you may get the first 200-500 KB super fast and then everything slows down.

If the website is resource heavy and has for example 100 pictures served from the same domain as the main page, you'll block your four connections downloading pipelined pictures and the page won't render because the browser is unable to create a 5th connection to get the js scripts or whatever else is required. Leave the value of max requests lower to 8-10, to force connections to be freed for such websites.

It's precisely due to websites that advertise this trick that sites like Google Maps has to resort to tricks like serving the pictures from several subdomains, so that each domain gets its own connections.
Otherwise, the 50-100 pictures on the page would get pipelined and loaded slowly by the browser.

Microsoft also does it :



By using ecn.dynamic.t0 , .t1 , .t2 , .t3 they go around the browsers' limitations of 2-4 connections per host and prevent pipelining, because it's important for each small image to be loaded fast, not to have 20 pictures requested on a single connection that may be throttled by the isp.

So.

* leave the caching to disk as it is, you only have to lose by caching to ramdisk or to ram. Windows caches in background to ram everything anyway.

* enable pipelining, it's good. BUT NOT ALL THE TIME, so compromise by setting a small maxrequests... 6-10 .. even 15 will be ok. 30 is often too much.

Too much pipelining only gives you the feeling the pages load faster. Instead, the page is blanck for longer time and then you get the page shown instantly, but if you time it, the pages may actually load completely in a longer time.

[pfrcom]

Originally Posted by mariushm

The value of 30 is too high

With clean installs of recent Firefox(es), including 18.0.1, network.http.pipelining.maxrequests is set to 32 by default

[TELVM]

Originally Posted by mariushm

... it's stupid to keep the cache in ram, if it was normal it would have been enabled by default ...

And so it is :

Browser.cache.memory.enable (default: true)

Browser.cache.memory.capacity (default: -1)

[mariushm]

Originally Posted by pfrcom

With clean installs of recent Firefox(es), including 18.0.1, network.http.pipelining.maxrequests is set to 32 by default

The 32 is basically a default value here, like "Unlimited", it's not an optimized value.
Remember, the about:config values are safe values, that must work for every computer in the world, they're not tuned for performance.

Originally Posted by TELVM

And so it is :

Browser.cache.memory.enable (default: true)

Browser.cache.memory.capacity (default: -1)

Have you read from the bug reports how this cache works?

This cache is different than the disk cache... it's used to hold in memory uncompressed gif icons, png icons decoded as bmp for faster rendering etc, and the default value is -1, which is sort of auto. As you close tabs, they get deleted from memory after some timeout value automatically.

It's not used like the disk cache is used, which holds files the website says that can be cached, between sessions and all that.

I incorrectly assumed from your device that it replaces the actual disk cache, which it doesn't. I never bothered to read about it before because simply from its name I know it's a bad idea.

And hint: there's also about:memory in firefox, which tells you a lot about where memory is used.