Firefox no longer starts?

[curaga]

Those default setting changes can make a difference in perception of speed - for example connections per server. You go to a news site, open a few tabs. Even if the total time to load them all is the same, if more tabs start loading instantly it looks better to the user.

[Jason W]

I see what you are saying though it is kind of like robbing Peter to pay Paul.  On a super fast connection it would improve the perception of performance; those with slower connections would probably prefer for a few tabs to load fully before more started loading.  But Firefox is obviously not made with slow machines or slow connections in mind.  I will try some such modifications and see if helps any.

In fact, limiting the default size of  that hideous urlclassifier3.sqlite file would be a modification that would help TC a lot.  I will rebuild and change that to about 15MB at most.  Those with persistent home can easily change that setting to be as big as they want and I will add that into the info file.

[tclfan]

I am definitely in favor of keeping Minefield available as the Firefox 3.0.x choice... I heard it once said that practicing security is more important than patching for security.  If you do not practice security but keep your system up to date, you may be at more risk than the user of older software releases that uses good security practices. 

Thanks Jason for this assessment... Is there Minefield 3.0.11 TCZ available? The current is 3.0.4...
In addition to your point on security we also have in TCL several mitigating factors:

- It is Linux after all... Focus for security threat is Windows as most common.
- TCL session runs as user, not root, which limits how much malware can do
- TCL does not suffer from 'System Rot' - refreshes itself on each boot to 'pristine' state, which limits exposure to malware.

Putting all this together, the window of opportunity for malware seems narrow... Minefield should do fine too. Opera is a great browser and very lean, but the lack of printing to PDF is the biggest problem.

[Jason W]

Rebuilt shiretoko and disabled the urlclassifier to eliminate that file altogether.  Default max size if nothing is specified was either unlimited or 104MB, either being unnacceptable.  Provided some other tweaks as well.  

The cache size is limited to 20mb now by default.  I ran shiretoko with youtube and surfed the web a while and the .mozilla directory never grows beyond 24MB altogether without that urlclassifier file.  When I exited the app the .mozilla directory shrank to 9MB.  It used to be common for me to see my .mozilla directory grow to over 150MB and that is dreadful on a live RAM based system.  Persistent home users can easily enough change the settings to store as much as they want.

[tclfan]

Thank you Jason for this. This is good news you eliminated that file. It was a cause of major headache for those appreciating efficiency...
Did you have a chance to do any other major optimization? What is your feeling of performance of this new Shiretoko, particularly compared to Minefield, i.e. memory and cpu utilization? Not quite benchmarks, but your opinion in TCL context?

[Jason W]

Basically these are the optimizations: i686 build with -O3 optimization, eliminated urlclassifier altogether, set default cache to 20MB, reduced maximum default popups.  The mozconfig file in the source directory lists the configure options for compiling.

I think that the difference in speed, memory usage, and such can be measured in a few percentage points between the official Firefox and a custom build.  Either way you slice it or dice it, it is still Firefox and is resource intensive.  Even the 2.0.0.20 build consumes almost the same memory as the current.  Of course, correct me if anyone has some benchmarks that tell otherwise.   But hopefully by at least eliminating the urlclassifier and trimming down default cache it will be more freindly for TC users, especially those that run with /home in RAM and back it up on shutdown.