[tor-talk] memory cached pages should reload instantly-but DON'T

[Joe Btfsplk]

On 5/7/2013 10:56 PM, David Vorick wrote:
> Are we sure this is a bug? Even when a page is in the cache doesn't it have
> to communicate with the server to verify that the cache hasn't expired?
> Perhaps this is what you are experiencing.
>
>
> On Tue, May 7, 2013 at 9:41 PM, Tom Ritter <tom at ritter.vg> wrote:
>
>> Hm, that's an tough question.  TBB doesn't modify the FF code very
>> much at all, and the patches are pretty lightweight - they're all
>> listed here:
>> https://gitweb.torproject.org/torbrowser.git/tree/HEAD:/src/current-patches/firefox
>> although some of them do deal with caching.
>>
>> The about:config settings are all listed here (AFAIK):
>>
>> https://gitweb.torproject.org/torbrowser.git/blob/HEAD:/build-scripts/config/pound_tor.js
>> so I wonder if there's anything in there you might recognize as
>> causing a problem?
>>
>> I'm afraid I'm not quite sure that the issue could be, these types of
>> bugs are pretty tricky to track down.  I did want to point you in the
>> right direction for maybe finding the culprit though.
>>
>>
I looked at torbrowser.git from link above.  One interesting thing. They 
show on line 16  pref("browser.cache.disk.enable", false); - but in my 
brand new extraction of TBB 2.3.25-6, in about:config, it has that as 
user set ENABLED.  But, I didn't enable it.  The last thing I'd want to 
do if looking for speed, is use disk cache. Didn't touch the setting in 
Options nor reset it in about:config; used a clean profile.  Others 
might check their setting in about:config.  This is kind of a big deal 
for not retaining TBB sessions on the machine.

In my TBB about:config, browser.cache.memory.enable is also = true (by 
default).  Only it has a much smaller max entry size than disk cache 
(which should be disabled): browser.cache.memory.max_entry_size;5120.
So default memory max entry size is apparently 5.1 MB (kinda small if 
you have 8 or more GB RAM).

I don't know what TBB is REALLY doing when I go back one page in 
history, but unless that page has expired, been purged from cache, etc., 
it shouldn't be reloading it.

Maybe it has to do w/ how long I use TBB in a session - I only just 
opened it today.  But today, going back a page or 2 in history is almost 
instant - as it should be.  Yet I've seen the slow issue I've described 
over many versions.  I hope it continues to make a liar out of me.

I'm not an expert on pipelining.  TBB sets 
network.http.pipelining.maxrequests = 12, though most documentation says 
#s higher than 8 will be treated as 8.

Later TBB versions have network.http.pipelining.aggressive =true, while 
it's disabled by default in Fx 20.  Some don't agree that it helps at 
all; in fact, just the opposite on some sites.
http://www.guypo.com/technical/http-pipelining-not-so-fast-nor-slow/
> The biggest impact, however, was the variance. Using aggressive 
> pipelining made certain sites MUCH slower. For example, about.com was 
> 2.5 times slower 
> <http://www.webpagetest.org/video/compare.php?tests=120725_BX_NF5-l:Pipelining+Off,120725_1J_NF4-l:Pipelining+On,120725_64_NF6-l:Aggressive+Pipelining> 
> with aggressive pipelining. Other sites balanced this effect by being 
> faster, with good results being 10-15% faster...