MozillaZine

Mozilla 0.6 released

Wednesday December 6th, 2000

mozilla.org today released a milestone based on the Netscape 6 branch. Builds are available on the ftp site, and more info on the build and it's purpose is available in the release notes.


#61 Re: Bloat solutions...

by LlelanD <LlelanD@TheSnakePitDev.com>

Monday December 11th, 2000 8:42 PM

You are replying to this message

My apologies for the length, but this issue was too important to allow it to be lost to the silly bickering (Of course, any developer who voluntarily works on an open-sourced project cares).

The underlying questions in this thread still remains unanswered: * Specifically, what is being considered or done to reduce the memory and OS resource consumption by the Mozilla browser. * If this information must be gained by following "Mozilla development", then at what sites is this specific information regularly available? * Is there a written history of the consumption numbers per platform for the releases/milestones to which we may refer?

I note that a single instance of build 2000120920 talkback zip the Sun JRE 1.3 plug-in on Win2K sp1 IE5.5 maximized on a 1152x864x32 display opened to <http://www.mozilla.org> has the following memory and OS resource consumption numbers via the Windows Task Manager (these number vary slightly each time): Physical Memory 23,532K Virtual Memory 19,992K Handles 301 USER Objects 49 GDI Objects 179

This looks pretty awful all by itself, but a sequence of operations proves enlightening. The rest of this is done in sequence.

A second identical instance shows these numbers: Physical Memory 26,020K Virtual Memory 22,468K Handles 300 USER Objects 43 GDI Objects 208 When you minimize the second instance (focus not on the first) you get: Physical Memory 4,068K // Slowly creeps up a little bit. Virtual Memory 22,512K Handles 300 USER Objects 60 GDI Objects 227 When you activate the first instance you get; Physical Memory 11,816K Virtual Memory 22,612K Handles 300 USER Objects 60 GDI Objects 227 When you minimize the first instance you get; Physical Memory 1,744K // Slowly creeps up a little bit. Virtual Memory 22,612K Handles 300 USER Objects 60 GDI Objects 227 When you re-maximize the first instance you get: Physical Memory 8,732K Virtual Memory 22,616K Handles 300 USER Objects 60 GDI Objects 227

The conclusion I can draw is that there has been significant work done on memory and OS resource consumption, but that a lot of it is not being activated or properly managed right off the bat. It takes some additional operations and events to fire up the new work. This sounds like a setup and initialization management problem, where things are being built and initialized correctly, but are then not being put into the properly shaved down state.

A comparison with MSIE also proves instructive.

MSIE under the same conditions: Physical Memory 10,724K Virtual Memory 3,340K Handles 268 USER Objects 117 GDI Objects 195 A second identical instance shows these numbers: Physical Memory 11,456K Virtual Memory 3,792K Handles 305 USER Objects 194 GDI Objects 263 When you minimize the second instance (focus not on the first) you get: Physical Memory 1,300K // Slowly creeps up a little bit. Virtual Memory 3,740K Handles 292 USER Objects 194 GDI Objects 265 When you activate the first instance you get; Physical Memory 3,020K Virtual Memory 3,740K Handles 292 USER Objects 194 GDI Objects 265 When you minimize the first instance you get; Physical Memory 796K Virtual Memory 3,740K Handles 292 USER Objects 194 GDI Objects 265 When you re-maximize the first instance you get: Physical Memory 2,548K Virtual Memory 3,740K Handles 292 USER Objects 194 GDI Objects 265

As you can see the OS resource numbers are comparable. The memory numbers may be lower, but this is not a fair comparison as a lot of IE is intrinsic to the OS. The numbers for the OS explorer process, which is heavily used by IE for UI, are as follows: Physical Memory 5,236K Virtual Memory 9,912K Handles 490 USER Objects 199 GDI Objects 408

These are about the same whether IE or Mozilla is up. Since Mozilla mostly uses its own code for the UI it obviously is going to be bigger than just the IE process. But when you combine the numbers for IE and the explorer process you get comparable physical memory numbers. This is only a rough reference since the explorer process is not dedicated to IE and serves the entire OS, but it does show that Mozilla is starting to get realistic numbers when everything is finally kicked in.

Mozilla can, has, and must do much better, especially when the first instance of the application is created since that is when most people will measure the consumption.

Providing findable, understandable, and consistently current information resources that tell us what has been, what is, and what is intended will go a long way in improving how this project is touted publicly and how we developers can help. I'm still burning most of my man-hours trying to learn this package on bad, outdated, overly-terse, and missing or unfindable docs. It's embarrassing.