Mozilla 1.0 Not Ready for Prime Time -- Close but No Cigar and No Brass Ring!
A Quick Look at Mozilla 1.0 Browser-Suite Performance -- Speed, Stability, and Memory Hogging
Mike Angelo -- 18 June 2002 (c) -- Page 2
There was no need to get out the stopwatch for these side-by-side tests. Mozilla was noticeably slower than both Microsoft Internet Explorer and NetCaptor. Moreover, our focus with this informal testing is the user-experience. If it takes a stopwatch to measure the performance differences, likely such difference will not have much of an impact on the user-experience.
Please keep in mind these tests were conducted on a 300-MHz machine. If your computer is faster than 300-MHz you might not notice any speed differences between Mozilla, MS Internet Explorer, and NetCaptor. When we ran these same tests on a 1-GHz box, the speed differences there were either not noticeable or not significant to the user-experience.
Just because performance differences between Mozilla, IE, and NetCaptor are not noticeable to the human eye on a high speed computer does not mean the reasons for those differences are not impacting overall system performance. For example, if those performance differences are due in part or in whole to such things as Mozilla taking more machine cycles and more memory than other browsers, then Mozilla is having an impact on overall system performance, even if it might only be a small impact. However, that discussion is beyond the scope of this article.
The big difference is not so much the Web-page display rendering-time. Mozilla is sluggish in all its operations on the 300-MHz test box. Likely, in part, this is because the Mozilla 1.0 browser is a set of scripts and HTML-based components running on top of the Mozilla Application Programming Framework. However, MS IE and NetCaptor are compiled binaries running directly on the Microsoft Windows OS without an application framework in-between -- sucking up resources.
There also are some deeply-rooted bugs that slow Mozilla performance down too, such as the stuttering bug -- Bug 91643. If you notice the Mozilla throbber on the Navigation bar freeze from time to time during page downloads, you likely are seeing the stuttering bug in action, which is an open bug at press time. Link in the Resources section on Page 3.
Overall, our browser side-by-side tests showed both Microsoft's Internet Explorer and Stylesoft's NetCaptor to be crispier and snappier than AOL-Netscape's Mozilla. Because Microsoft Internet Explorer and Stylesoft NetCaptor do not have e-mail client modules, there obviously are no side-by-side comparisons to make e-mail-wise.
However, the Mozilla e-mail client on the 300-MHz box is intolerably slow. You might not notice that sluggishness if you have only one mail account or you use a fast computer. But, if you have several e-mail accounts and often switch from one mailbox to another, you will find yourself doing lots of thumb-twiddling and time-wasting waiting for Mozilla to switch from one mailbox to another, and from one message to another, on a not-so-fast machine.
When composing e-mail on the 300-MHz machine, we find it faster to compose the e-mail in AbiWord and then paste the message into the Mozilla e-mail composer. When replying to e-mail with in-line responses, it is faster to cut and copy the text in the Mozilla e-mail composer panel, paste it into AbiWord, add the comments/responses, and then copy and paste it back into the Mozilla e-mail composer panel. Additionally, AbiWord does spell checking -- the Mozilla 1.0 e-mail editor does not!
You should not have to leave the e-mail composer for any of these reasons. This sort of stuff is yet another reason why Mozilla 1.0 is premature and not ready for prime time.
Next, the 31 May 2002 Mozilla 1.0 Windows test build or 5 June final Mozilla 1.0 build, MS Internet Explorer, and NetCaptor were opened with blank pages in the Web-page display panels. That brought the Free Resources down to 56%.
In the side-by-side Web-page Display from Cache test, each of the three browsers was taken from the blank page to the MozillaQuest Magazine Front Page (MozillaQuest.com). That served to place the page in each browser's Web page cache. Then the Mozilla 1.0 is Officially Out! article was opened in each browser.
Next, the back button on each browser was hit in quick sequence starting with the Mozilla browser. Both Microsoft IE and Stylesoft NetCaptor completed their display work before AOL-Netscape's Mozilla was done -- by several seconds.
However, in a few trials of Display from Cache tests run on a 1-GHz Windows 98 SE computer it appeared that Mozilla might have been just a fraction of a second faster than Internet Explorer. Nevertheless, on most trials of the Display from Cache tests run on the 1-GHz Windows 98 SE computer there was no noticeable speed difference.
Please note that Web-page Display from Cache tests do not directly measure performance of the rendering engine that a browser uses. That's because the user interface (UI) and other browser components are involved in the process too. A slow UI or other browser components could slow down the time it takes to display a Web page even if the underlying rendering engine is fast.
The Mozilla and Netscape people claim the Gecko layout/rendering engine, which both Mozilla 1.0 and Netscape 6/7 use, is very fast. Yet our side-by-side Display from Cache tests show Mozilla to be slower than Internet Explorer and NetCaptor. Both IE and NetCaptor use the Internet Explorer rendering engine.
If the Mozilla and Netscape claims that the Gecko layout/rendering engine is so fast are true, but yet Display from Cache tests show Mozilla to be slower than Internet Explorer and NetCaptor, then that suggests that the bottleneck is in the Mozilla UI, other browser components, and/or the Mozilla Application Programming Framework.
In these side-by-side tests, the task at hand was started on Mozilla first and then immediately started on the other browsers in quick sequence. The side-by-side test results are summarized in Table 1, below.
The next test was to compare how quickly each browser could be brought from minimized status to focus on the desktop. This was done by minimizing all windows so that only the Windows desktop and Taskbar were visible. Then each browser's icon on the Windows Task Bar was clicked in immediate sequence, starting with Mozilla.
The results of the side-by-side Up from Task Bar tests produced mixed results. On some occasions both MS IE and NetCaptor beat Mozilla by about a second. On other occasions, the speed differences were hardly noticeable or not noticeable at all.
Here the sidebars were closed on all three browsers. Then the sidebars on the Mozilla and IE browsers were opened in immediate sequence starting with Mozilla followed by IE. Next the Mozilla and NetCaptor sidebars were opened in immediate sequence starting with Mozilla followed by NetCaptor. In these tests, the Internet Explorer and NetCaptor sidebars opened noticeably faster than did the Mozilla sidebar.
In similar fashion Sidebar Closing tests were run. There were no easily noticeable differences among the three browsers in how fast the sidebars closed.
Next side-by-side tests involving increasing and decreasing text size were conducted. Again, Internet Explorer and NetCaptor were faster than Mozilla. Interestingly, IE and NetCaptor accomplished the text size changes even before the Mozilla View > Text Size drop-down menu collapsed after clicking on the selected text size.
In the side-by-side Open Help > About tests, IE and NetCaptor again were noticeably faster than Mozilla. Incidentally, both Internet Explorer and NetCaptor responded to depressing the F1 key by popping up their in-program Help panels. Mozilla does not seem to do this. (Interestingly, the F1 trick works with Netscape 7.0-PR1 which is based on the Mozilla 1.0-RC2 code.)
The results of the side-by-side Close Help > About tests produced mixed results. On some occasions both MS IE and NetCaptor beat Mozilla by about a second. On other occasions, the speed differences were hardly noticeable or not noticeable at all.
A side-by-side test run that is not shown in Table 1, is the Open Web Page from Local File test. In this test, the File > Open File menu was popped up and the MozillaQuest Magazine Front Page URL on our LAN (Local Area Network) was cranked into the insert box in the Open File panel in each of the three subject test browsers.
Opening a Web page from a local network location rather than from an Internet location is an important test because there are no extraneous influences due to Internet connection speeds, bandwidth problems, traffic loads, and so forth. Pretty much all the time it takes to get the page displayed is due to browser efficiency, or lack thereof.Overall, our side-by-side comparisons appear to indicate that the AOL-Netscape Mozilla browser is noticeably slower than both the Microsoft Internet explorer and Stylesoft NetCaptor browsers on our 300-MHz Windows 98 SE test machine.
You should not have to put up with sluggish performance. Nor should you have to buy a new, very high-speed computer just to use an inefficient browser-suite These sluggish performance problems are yet another reason why Mozilla 1.0 is premature and not ready for prime time.
However, CNET found Mozilla faster than Internet Explorer in three of four Web-page loading tests it conducted using Windows XP. Although, CNET reported the actual loading time differences were less than a second. From the CNET report it appears that CNET pretty much employed the same general sort of procedure we did in our Display from Cache test.
Interestingly, Mozilla loaded pages more slowly than did Netscape 6.2 in three of the four CNET page-loading speed tests. Could this mean that the Netscape 6.2 UI is faster than the Mozilla 1.0 user interface?
Netscape 6.2 is based upon Mozilla 0.9.4 code whereas Netscape 7.0-PR1 is based upon Mozilla 1.0 release candidate 2 (RC2) code. Unfortunately, CNET did not also report comparable test data for Netscape 7.0-PR1 and Mozilla 0.9.4. That makes it difficult to make inferences from the CNET tests as to whether the speed-performance differences between the Netscape and Mozilla browsers result from changes in the underlying Gecko rendering/layout engine from the Mozilla 0.9.4 code-base to the Mozilla 1.0 code-base -- or from differences in the user interfaces.
Copyright 2000-2002 -- MozillaQuest -- Brodheadsville, Pa..USA -- All Rights Reserved