Tuesday September 18th, 2001
We've got a new poll up, for your polling pleasure. It deals with the new Quick Launch feature, and we wanted to know what you think of it. If you think it's buggy, or have improvements, you can note that here, or in the poll's talkback. Poll away!
#1 I've been using quicklaunch in Netscape 6.1
Tuesday September 18th, 2001 1:26 PM
Netscape 6.1 shipped with Quick Launch ages ago. I believe that was based on 0.9.2.
#2 Re: I've been using quicklaunch in Netscape 6.1
Tuesday September 18th, 2001 3:56 PM
Yes, but it's been improved since then. And of course it's enabled by default in the 0.9.4 installer now (much to the horror of MozillaQuest!).
-turbo is a disgusting hack. What's the pratical difference between -turbo and just minimizing your mozilla windows?
People care a lot about the time their browser / wordprocessor / instant messaging application takes to start up and be usable. That's one of the main reasons people keep using Internet Explorer over Netscape (or Mozilla which is less known to this kind of users). This is far better than minimizing a window because it gives the sensation that the mozilla browser itself is speedier (which actually is non related). As I said in the beginning "It's all about the feeling" (for not too advanced users anyway). Of course, just my .02
#7 Re: It's all about the feeling
Tuesday September 18th, 2001 7:07 PM
So what about the new feeling that people with 32-64 MB of memory will get after installing Mozilla (or Netscape 6.x where x > 1) when they discover that ever since installing Mozilla/Netscape, EVERYTHING on their computer runs slower (due to less avail RAM, more virtual-memory swapping, but they don't realize this). Not the sort of feeling we really want to be spreading around. Not for the sake of trying to fool users in the same evil way that Microsoft does with IE, and give them the wrong impression (not to mention leading programmers down a very bad path for future applications).
-turbo should be an option, not default, and only after being presented with a clear explaination of what happens when you enable it and what negative side-effects can result. I certainly know _I'll_ never use it.
#8 Re: Re: It's all about the feeling
Tuesday September 18th, 2001 7:31 PM
It is a option in Netscape 6.1 and it was not enabled by default until recently. Even then there is a whole section in the installer that describes it so I dont think anyone will not realize that its on.
#11 Re: Re: It's all about the feeling
by trudelle <email@example.com>
Tuesday September 18th, 2001 10:58 PM
Have you experienced such a slowdown yourself, or seen it happen on someone else's machine? What other negative side effects have you seen result? Why are you certain you'll never use it?
#21 Re: Re: Re: It's all about the feeling
Wednesday September 19th, 2001 6:16 AM
I see this slowdown on a 64MB/500MHZ machine, after some using of Mozilla (monitoring allocated memory I see it always increasing after Mozilla operations). Slowdown is evident when I open memory demanding applications (Acrobat, IE). A lot of disk activity (because of memory swapping). Admittedly, not to the point of making system unusable but yes it's annoying.
#25 Re: Re: Re: Re: It's all about the feeling
by arnoudb <firstname.lastname@example.org>
Wednesday September 19th, 2001 7:34 AM
Then buy some more RAM. It's not like it actually costs anything these days. There's just no excuse to have less than 256 with these prices, no matter how tight your budget. Having only 64 megs is hardly enough to run Mozilla anyway, -turbo or not. If you have plenty of RAM (I have 1GB), then -turbo is definately something you'll want, it eats many seconds of Moz's startup time, even on an Athlon 1200.
#33 Other plattforms + minimum ram requirement (64MB)
Wednesday September 19th, 2001 9:13 AM
PDAs or other special computers cannot be equipped with such huge amounts of RAM. Regarding to PCs, you can't easily convince your employer to upgrade corporate PCs (like this one with the 64MB I am using at work) just to be able to run Mozilla.
I do agree that footprint will become less important in the future but 64MB is still the minimum requirement for Mozilla. Unless they change it, current footprint is unacceptably large, imho.
#42 Re: Re: Re: Re: Re: It's all about the feeling
Wednesday September 19th, 2001 1:23 PM
Telling people to spend money to use Mozilla is not going to work.
#45 Re: Re: Re: Re: Re: Re: It's all about the feeling
Wednesday September 19th, 2001 4:16 PM
If by 'work' you mean piss them off, then yes, it probably will work ;)
Mozilla likes RAM. Telling people that Mozilla likes RAM seems reasonable. Telling people who have machines that don't run current Mozilla well that they should use something else or upgrade seems like a reasonable thing to do. Telling people that Mozilla is acceptable to some other people that have more modern equipment also seems reasonable. I happen to have machines that span about 5 years and I use Mozilla on all of them. It's pretty slow on my P100 with 32 MB ram, it's better on my PPro 200 with 64 MB RAM, it's even better on my PIII 500 with 128 MB RAM and it works great on my PIII 650 with 256 MB RAM. I also have an old Power Computing 120 with 80 MB RAM and it's a bit slow there but absolutely fine on my G4 450 with 356 MB RAM.
Mozilla likes the RAM. It takes some RAM and some disk footprint to support as much as Mozilla does. It continues to get better and I think that I'll be completely comfortable with Mozilla on 64 MB within the next 6 months, especially with quick launch. Quick launch is great for me _especially_ on the low end machine. The browser is my primary app and the one place that it's painful is on startup. I don't have any problems running a word processor or a 2D game like chess at the same time and that's about all I need. For many people out there who use Netscape and other vendor releases based on Mozilla I expect that experiences will be the same. For "power users" I don't recommend quick launch, nor do I recommend doing much of anything on 64 MB machines. If you really love your 64MB machine or don't have the money or motherboard for an upgrade of cheep RAM and Mozilla/Netscape/other isn't acceptable for you then use something else and check back every so often to see if it's gotten any better
BTW: According to C|Net, which is usually partial to IE, Netscape QuickLaunch added an avereage of only 4 seconds to startup, and they noticed no difference.
#53 Re: Re: Re: Re: Re: Re: Re: It's all about the fee
Wednesday September 19th, 2001 9:11 PM
By "work," I mean, "convince people to use Mozilla or one of its derivatives." I agree that it is reasonable to say people should upgrade their hardware or use a non-Mozilla web browser. That will not "work," though.
> It continues to get better and I think that I'll be completely comfortable with Mozilla on 64 MB within the next 6 months, especially with quick launch.
I like that. But I think there should be more memory usage specifications than a simple 64MB minimum hardware limit. E.g. a maximum footprint after a number of specific operations. And footprint problems don't seem to have any priority for the team working on performance issues (see performance meetings), afaik.
A query on Bugzilla, see <http://bugzilla.mozilla.o…se+same+sort+as+last+time> , (using "footprint" and "mlk" keywords) reveals 189 open bugs. Less than half of them are assigned. People who are interested in memory usage problems might contribute in any technical aspect of these bugs or simply vote.
If you do your query on keywords crash and topcrash you'll get about 439 open bugs, and of these are only 118 assigned. On the other hand we surely know that stability has high priority for developers, so I think this shows that the percentage of assigned bugs may not be the best means of seeing priorities...
#52 Re: Re: Re: Re: Re: It's all about the feeling
Wednesday September 19th, 2001 8:21 PM
Telling everyone to upgrade to 1GB of RAM just because it's cheap isn't a legit reason for letting a browser suck up far more RAM than it needs to.
As it is, Mozilla without -turbo uses too much RAM (I know, it's being worked on).
Most users will never upgrade parts of their computer. And I don't like to see my favorite software project cave in to the bad industry practices of allowing bloat because RAM/HDD space/CPU MHz is cheap. People with 64 MB of RAM who are perfectly happy with the computer speed NOW and want nothing more than a reasonably-performing IE-alternative need to be catered to. This is not too much to ask. -turbo (if enabled by default, which it sounds like it might not be, thank goodness) works against that idea.
I have no problem with -turbo being an OPTION ( I AM a champion of having choices, and that works both ways), but I'm against having it as ON by DEFAULT (except for the temporary need to get a wider audience testing it). As long as Netscape leaves it OFF, I'll be happy.
#58 Re: Re: Re: Re: Re: Re: It's all about the feeling
Wednesday September 19th, 2001 11:21 PM
It is ALWAYS an option. You are presented with a question in the installer. You can check or uncheck the checkbox and hit OK. The box will either have a check already in it (easy enough to uncheck) or it will not have a check in it (easy enough to check). If that wasn't enough the feature can easily be enabled or disabled in the preferences. Either you don't understand what enabled by defauly means or you think that users that care about memory usage are too stupid to uncheck a checkbox or disable a feature from the preferences or the taskbar. Am I missing something? Has this not been explained well? Have you used a Mozilla or Netscape installer?
#78 Re: Re: Re: Re: Re: Re: It's all about the feeling
by arnoudb <email@example.com>
Sunday September 23rd, 2001 1:03 PM
I still really don't see the problem with it being on by default. Mozilla is *not* an end-user product, no matter how many of us would like it to be that. Normal users have never even heard of Mozilla, Mozilla is for developers and interested power-users. I don't think either of these two groups will be unable to check or uncheck a checkbox. As for unwanted or "politically-incorrect" settings being forces on you, Mozilla has thousands of prefs, most of them hidden, but you just can't please everyone, that's why we even have all those prefs. That's also why the prefs are set to which might please *most* users. If you have a problem with a setting, just disable it, it ain't that hard.. Try installing RealMedia Player. Makes the problem of unwanted prefs in Mozilla seems smaller suddenly, eh?
#51 Re: Re: Re: It's all about the feeling
Wednesday September 19th, 2001 8:16 PM
I have 512MB of RAM on my own machine at home. At work, I have 128 (soon to be 256).
But I know plenty of people with 64MB of RAM who hate IE and want an alternative. Hell, even 128MB isn't enough once you get Windoze loaded. Then there's AV software. Plus all the crap PCs come with these days that load in the System Tray and take up dozens of MB of RAM at all times on the off-chance that someday the user might use it once or twice. Sure, you and I know how to whack this stuff, but not the average user.
Telling everyone to upgrade the RAM in their PC isn't appropriate either. Most PCs don't get upgraded... the idea is either foreign or scary to too many users. They remain configured as-purchased. Users are FAR more likely to download and install new software (like Mozilla/NS6.x). As long as -turbo remains an option you have to ENABLE, I have no problem (beyond the fact that even without -turbo, the memory usage requirements and UI speed of Mozilla need serious improvement).
#12 Re: Re: It's all about the feeling
Tuesday September 18th, 2001 11:01 PM
I can't say what Netscape plans are beyond what they've done in Mozilla and this feature has always been and as far as I can tell will always be an option. There are two places to opt in for the feature and two places to opt out fot the feature. You have to say yes in the installer to turn it on and if you say no there you can turn it on in preferences. You can turn it off in preferences or by context-clicking on the tray icon and disabling quick launch. No one is trying to fool anyone. A user says "yes, I'd like mozilla to stay resident in memory so it starts faster". What's fooling there. The other guys don't give users any choice. I don't like that.
I happen to really like this feature. I don't use it because it can get in the way of testing Mozilla builds but if I was a consumer with a contemporary computer (something that was new in the last year or two)using a Netscape release then I'd definitely enable this feature.
#24 Re: Re: It's all about the feeling
Wednesday September 19th, 2001 6:41 AM
What on earth is this bizarre line about fooling users. I know what -turbo is. I know what it does. And in the rare occasions that I boot into Windows, I have it turned on. Why? Because I want my mozilla browser to appear a couple of seconds after my wanting to first use it, not ten... I'm under no illusions, I know how it work: and it works as advertised, my browser appears more quickly when I want this. This is 'disguising the fact that the browser appears just as slowly' - I don't care how long it takes to load, I care how long it takes to appear when I click the icon.... Where's the confusion?
#10 I hate MS
Tuesday September 18th, 2001 8:23 PM
Plus, it doesn't clutter up on the taskbar. Very important for me; I don't want my taskbar to get filled up with programs I don't actually _use_ at the time. If I need the system ressources Mozilla takes up, I'll just exit the quicklaunch fucker.
#13 Re: Nasty hack
by trudelle <firstname.lastname@example.org>
Tuesday September 18th, 2001 11:04 PM
The new implementation handles multiple profiles correctly, and really terminates your session,logging you out of server connections, etc. Why do you think it is a 'disgusting hack', any more than say, a pre-fetching cache is?
#15 Another difference between -turbo and minimize
Wednesday September 19th, 2001 12:35 AM
The difference is user might accidentially exited the application (mozilla) by closing the last browser window. Isn't it a hassle if you need to check how many window left when deciding whether to close or to minimize a window?
Don't know how many of you were DOS users (or even Win3x users with limited memory systems). An ex-DOS user has the habit to close everything he doesn't use at the moment, even if he will need it in the next two minutes. I always try to resist this tendency to prematurely close my browser, but many times I fail.
#4 Limited Poll choices
by SubtleRebel <email@example.com>
Tuesday September 18th, 2001 5:05 PM
Those of who use Linux, Solaris, and MacOS instead of Windows do not have much choice but to choose "No" and thus I think the results could be misleading.
they could say "Who is Turbo?" like one did....but the results are still the same...
#9 I hate MS
Tuesday September 18th, 2001 8:17 PM
Except for the quicklaunch feature just becoming unresponsive after the first run of Mozilla, it's been working perfectly so far (several minutes). Oh, and the icon has been poorly converted to taskbar size.
But love it already! Hated those long loads for old versions of mozilla, made me uninstall in favor of MSIE!
#16 Quicklaunch is depended on footprint/leak problems
Wednesday September 19th, 2001 1:48 AM
Hence the need for reducing them asap. Any system with less than 128 MB RAM is unable to use quicklaunch without severe performance degradation due to swapping. Creating a mail message, reading news or opening a few browser windows increase footprint to unacceptable levels (for a 64, even a 128MB machine). For Win9x, it's worse than that: system might become unstable at the threshold of swapping (probably this is bad OS memory management). Nevertheless, quicklaunch is a great feature. I expect it to become a cross-platform feature.
#36 Re: Quicklaunch is depended on footprint/leak prob
Wednesday September 19th, 2001 10:11 AM
It is already in a way - MacOS keeps all applications in memory until you tell them to quit. No linux yet of course.
I just wrote the post below to the poll talkback, but I figured it would probably be more visible here so I'll re-post it here:
I'm curious as to where this "IE is preloaded" thing comes from. Sure, IE really does load damn fast so you'd think there was some (partial?) preloading or similar technique used but does anyone have any proof of anything one way or the other or is this just FUD by the anti-Microsoft league? Has Microsoft said anywhere that they pre-load? Does anyone know what parts of IE that are loaded? What DLL's? Are these DLL's also used by other parts of the system anyway? Any metrics on this - not just rumors and speculation?
The fact that parts of Windows Explorer share the same UI (eg menu list) could be an indication. The way OS handles IE updates (system restart needed etc) might be another one as well as the ridiculously low footprint indications on TaskManager. Surely I am not the person suitable for giving a solid proof but others in mozilla-performance newsgroup are able to do this. It would be very interesting to see such an analysis, to the point that it would deserve a dedicated mozillazine article.
The UI of IE is pre loaded? Oh my god! That must be several kilobytes worth of code! Come on.. get serious. If IE *IS* pre loaded (which it very well might be, that's what I'm asking here), then the UI probably is one of the few parts that are NOT pre loaded.
"The way OS handles IE updates (system restart needed etc) might be another one".
Umm, so does a million other pieces of software. The reason for this is that some DLL's that IE uses (and the other software that requires a reboot) can't be unloaded and thus requires a reboot at which point the DLL's are replaced by the new versions. Many games requie a reboot. So does the Java VM. That's not a proof of preloading though.
"ridiculously low footprint indications on TaskManager."
I don't know about ridiculously low. You can check the exact amount of memory each process uses with, for example, the process viewer tool that comes with Visual C++ / Visual Studio. It shows every loaded DLL, and how much memory is used and where that memory goes. The IE process uses about the same amount of memory as Netscape 4.. An empty IE window uses about 7 MB.
Proposing three tests, the latter being not so scientific but still better than the other two: 1. Find which dlls are used by both IE and Windows Explorer and they weren't included in previous versions of Windows (such as common dialog controls etc). The latter might be a difficult task.2. Find which dlls used by IE are memory resident after Windows startup. 3. Take a Win95A system without IE installed and measure startup time. Then, install IE 5.5 on it and measure system startup time again. I bet you will see a significant increase, particularly if you don't have a state of the art hard disk subsystem.
I 'll try the second test and post the results here. Unfortunately, I don't have a clean Win95A system available. If someone has, please do test 3.As for your objections about UI preloading, I am still not completely convinced. You can't be sure how much code you need for maintaing its caches etc. I am just a hobbyist programmer, yet my instinct says it can't be attained as easily as you say. Still not an evidence, as I told you before.
As I already said, the whole system-restart-before-use thing is NOT an evidence. But we have Mozilla (as well as NS4x and many games) as a proof that some complex applications *do not* need restart to work. Btw, why IE needs to unload some dlls ? When I update it from local installer, IE is supposedly *not* resident. But let's consider installer contains critical system files (which, strangely enough, have not been published in other system updates), that they need update. Isn't this a strong indication that those "common" files are part of the preloaded IE ?
#29 Re: Re: Re: Re: IE preloading
Wednesday September 19th, 2001 8:29 AM
Oh, definitely! DLL's that IE use are also used by other parts of the system. I'm not denying that IE is preloading. I'm wondering to what extent it's preloaded or if "preloading" is even the correct term in the case of what IE is doing.
#32 So, it's a matter of terminology
Wednesday September 19th, 2001 8:52 AM
Your question about the percentage of "preload" (or whatever we want to call it) is interesting. Let's do some testing to reveal this, then. I think it worths the effort.
#37 Re: Re: Re: Re: Re: IE preloading
Wednesday September 19th, 2001 10:16 AM
I used Norton's System Information to see what DLLs are used by explorer.exe and iexplore.exe and NOT used by mozilla.exe (because Mozilla is the same kind of application as Internet Explorer, I thought that would be the most useful means of distinguishing them). Unfortunately, I managed to crash my system in the process and lost the list, and I don't have enough time to do it again; but it comes down to a large list of DLLs used by iexplore and explorer and not used by mozilla, most of them used only a few other programs (for instance, Office 2000 requires the use of a lot of the functionality in IE5+), and a good-sized list of DLLs used by all three (and therefore reasonably considered to be "operating system" DLLs, depending on how you want to define them), and one DLL used by explorer and mozilla and not used by iexplore.
Mozilla uses 8.3 MB in the private modules, and IE 1.97 MB in the private modules; while the total sizes are 12.9 MB for Mozilla and 14.7 for IE (much to my amazement). This is with each browser having only 1 window open on a web page (for mozilla, this page; for ie, the google home page), no turbo loaded, ie6 versus mozilla 2001081703, Windows 98 SE.
Hope this answers your questions, macpeep.
The file manager(Windows Explorer) is IE. Don't believe me type in <http://www.google.com> into the location bar of your My Computer Directory Window.
You just proved (well, not really you didn't, but I know this from before) that explorer.exe is IE. You didn't prove that IE preloads. *PRELOADS*. Everyone is saying stuff like "uninstalling IE frees 30 MB of disk space" etc. but nobody is answering the actual question: how much and how does IE preload? In other words, how much of IE is actually loaded at startup and in what form? Shared DLL's? Running, but with no visibile windows? etc.. We all know that IE totally integrated with the OS and all that. That's not the point.
I'm confused by your confusion. You are aware that explorer runs as soon as Windows boots, aren't you? That's what your desktop is. IE = explorer, and explorer runs when you boot, so it preloads practically the whole of IE, excepting perhaps the back and forward buttons and spinning e....
Well, not necessarily. We don't know what parts of "IE" explorer uses. If it runs in its own process completely, it is possible that the IE that you launch isn't actually preloaded at all. Just because the OS uses IE doesn't necessarily mean that the IE processes that you launch to surf are preloaded in any way. That's not the same thing.
I think tny's numbers show something but we need further investigation (hoping I post something tonight). His report about private memory usage support my opinion that TaskManager (or other similar utilities) reports ridiculously low memory numbers for IE. Just 1.97 MB of allocated private memory ! Is it possible ?
Erm. The simple answer is that yes, Microsoft have said that they did this. This is what it being integrated with the OS is all about. Lots of what you see in Windows (98 and beyond) is produced by the Internet Explorer engine. Your desktop is, as can be seen very clearly if you turn Active Desktop on, Windows Explorer is, help files are... There has been a whole lawsuit about them tying IE to the OS - and this is when IE is loaded into memory - when you start using these contentious parts (i.e. upon boot)...
#26 Re: IE preloading
by arnoudb <firstname.lastname@example.org>
Wednesday September 19th, 2001 7:45 AM
Not to mention a crashed Internet Explorer taking down the whole system on Win9x platforms.
And what do you think this whole antitrust lawsuit is all about?
"Not to mention a crashed Internet Explorer taking down the whole system on Win9x platforms."
I guess Quake III is also part of the Win9x platform then...
"And what do you think this whole antitrust lawsuit is all about?"
It was about bundling - not PRELOADING.
#34 Re: Re: Re: IE preloading
Wednesday September 19th, 2001 9:19 AM
Macpeep said: > It was about bundling - not PRELOADING.
Yes, and Microsoft's defence was 'We had to bundle it, it's inextricably tied into the operating system, even Windows Explorer uses it'...
#35 Re: Re: Re: Re: IE preloading
Wednesday September 19th, 2001 9:53 AM
Yes, and that says nothing about preloading - it says something about dependencies.
#39 Windows without preloading IE
by SubtleRebel <email@example.com>
Wednesday September 19th, 2001 12:53 PM
If you are really interested in finding out how much is preloaded for IE when Windows boots up, I suggest you look into Win98Lite and IEradicator.
These programs will allow you to remove IE integration from Windows 9x. So all you have to do to find out what is preloaded for IE is to check open DLL, memory usage, etc on a typical Windows boot up and then use one of these programs to remove IE, reboot, and check the DLLs, memory, etc again.
I skimmed through the website to see if it said anywhere how much memory was freed up or what the relevant DLLs were, but I did not find it. I did find some benchmarks comparing running various apps on Windows with IE preloading to the same apps under Windows without IE. It also said at one point that deleting IE from Windows provided an additional 40+ MB of hard drive space.
We know IE is preloaded because Microsoft tells us that it is (they would not lie about that would they?), but even if MS did not admit it, the evidence is easy to find. Suggesting that IE does not preload is like suggesting that Windows 9x/ME does not rely on DOS.
#41 Re: Windows without preloading IE
by SubtleRebel <firstname.lastname@example.org>
Wednesday September 19th, 2001 1:06 PM
Looking at info on IEradicator on a different page of the same site, it says that removing IE will clear off 30+ MB of disk space rather than the 40+ MB that I saw on a different page. I am not sure which number is right, but both came from the 98lite.net website.
BTW,IEradicator is a free download and works with all Windows 95/98/ME and IE 3 through IE 6.0 beta 1 (although not all of them preload) --- perhaps the difference between the 30+ MB and 40+ MB depends on which versions of Windows and IE are involved.
For more details on what all IEradicator does, see <http://www.98lite.net/ierad.txt>
I can think of a couple instances where IE is loaded on bootup. Active Desktop is IE on your desktop and the web view in explorer windows. The search sidebar under win2k uses web pages (and you can't disable that). Under win2k the recycle bin shows form buttons.
I'm sure there are other places where it's used as well.
#18 Installer should check available RAM
Wednesday September 19th, 2001 5:32 AM
Bug 100516 - I really think the installer should check the amount of physical RAM in a system before defaulting the installer to on.
This means people with a 64MB machine who just blindly accept the defaults will not have the preloader enabled, but those with more memory (e.g. 128MB) will. It's not an ideal solution but better than nothing.
If you'd like to see this in the installer then vote for this bug (no 'me too' comments tho please), however if you've got any useful comments like what you think the minimum RAM to default preloader to on should be, then add them to the bug.
Provocative title, I know.. :) What I mean is this:
Suppose you have 128 megs of RAM, use Mozilla with -turbo and thus have it preloaded. Now suppose you just use your machine for other stuff for a few hours.. play games, use Star Office and other stuff that uses a lot of RAM. If I understand correctly, the Mozilla in the background is more or less completely idle and will be swapped out to the disk.
Now if you "launch Mozilla", how much faster will the launch actually be? The preloaded Mozilla will swap out from the disk into RAM instead of just loading form the disk. Will this actually be any faster? I can definitely see that loading, exiting, loading, exiting would be way faster with -turbo mode, but that's not exactly a typical situation..
#55 Re: Does preloading work?
Wednesday September 19th, 2001 11:06 PM
If people load and exit over and over then the browser will be preloaded anyway. I mentioned the temporary nature of the preloader's effect. Somebody told me something absolutely stupid and completely ignored the point that the preloader is not a huge benefit.
#76 Re: Re: Does preloading work?
by jesse <email@example.com>
Saturday September 22nd, 2001 9:39 PM
For me, Quick Launch makes (exit browser, relaunch browser) *much* faster. I have Win98, 400mhz, 128mb ram, 09/22 nightly. (Note that in 0.9.4 and builds right around the milestone, Quick Launch didn't have much effect all).
What setup are you using?
#59 Re: Does preloading work?
Wednesday September 19th, 2001 11:36 PM
Great Question!! How about playing around with Mozilla and a stopwatch. Report your findings back here. That would be great. Thanks.
#63 Re: Re: Does preloading work?
Thursday September 20th, 2001 1:54 AM
Ok.. I did.. I ran the stuff on my computer at work; HP Vectra, 800-something MHz, 256MB of ram, Windows 2000. The tests were made with build 2001091903, which was the latest nightly build at the time of the tests.
First I installed Mozilla and configured it the way I want it (theme, bookmarks, toolbar buttons, sidebar panels etc.). Before installing, I deleted any trace of any old Mozilla's (All files from my "documents and settings\username\application data\mozilla" dir, as well as all files from the old Mozilla "bin" dir. Did I miss something btw?).
Then I enabled QuickLaunch and tried it out a few times.. And finally exited so that only QuickLaunch was running in the taskbar tray, but no windows open. I then loaded Visual C++ and compiled an app that took about 10 minutes or so to compile and heavily trashed the hard drive and hopefully swapped out everything from RAM. I then proceeded to open a new window from the Mozilla quick launch tray icon and timed it. Result: 5 seconds.
I exited Visual C++, waited for things to calm down a little, closed Mozilla and then relaunched a Mozilla window from the tray icon again. Result: 3 seconds.
Finally, I disabled quick launch, waited for about 30 seconds and ran Mozilla. Result: 5 seconds.
Not very conclusive results I'd say, since all the loading times are relatively low. This is why I'm not that excited about Quick Launch myself. The loading time is fast enough for me anyway. At home, I have 384 MB of RAM on my desktop machine so there's a considerable disk cache anyway. In any case, the little this test showed was that a swapped out quick launch isn't much slower than a non swapped out, but then again, it's also not any faster than a normal load.
I'd like to see what kind of results other people are getting on a similar test..
#65 There's a bug that distorts your tests
Thursday September 20th, 2001 3:07 AM
This bug <http://bugzilla.mozilla.org/show_bug.cgi?id=99387> has almost elliminated -turbo mode (quicklaunch) gains. At this moment, quicklaunch startup time is almost identical to the conventional relaunch time, if there's adequate ram for disk caching. Unfortunately, your tests are that useful until that bug will be fixed.
#69 Re: There's a bug that distorts your tests
Thursday September 20th, 2001 1:58 PM
If this is true then people should stop bragging about the quick launch thing until it works.
You' re not right. Preloading isn't just as simple as it sounds. If you check the quicklaunch Bugzilla tracking bug, you 'll see a lot of bugs fixed and a few ones still open. Quicklaunch needs testing because it's so complicated task. Btw, they already checked in the tree a fix for the performance bug we discuss. Builds using this fix will be available today.
I've no idea about how -turbo actually works so this might not make sense at all:
Instead of having just a check box, what about we use a slider to choose how much to preload? I think many systems with moderate amont of memory would benefit from being able to preload 5MB instead of 20MB.
I think at least 3 states for preload nothing, preload only the most time consuming parts wich don't take so much RAM and preload all.
But I don't know if there are such parts that take long to load but don't take this much RAM (maybe parsing and loading the prefs?)
#46 Should become useful when idle
Wednesday September 19th, 2001 4:33 PM
I have wrote this before. The taskbar icon needs to become something useful, something people want to have even if they are not using mozilla. For example:
- Users should be notified visually about a new email, moreover about a new email from which account. (There might be an option about which email accounts should be used in taskbar alert).
- This one is related mostly with netscape. AOL instant messenger alerts. I want to see them. I want to be connected to AIM and I want to be alerted when there is new buddy online, etc.
- I want to be notified if a webpage is updated. For example slashdot's main page. So I need to be able to register some websites to be hecked for updates and I need a not-so-disturbing alert for this.
Once you transform the taskbar icon into something useful, then nobody will complain about how much memory it takes and all...
#48 Re: Should become useful when idle
Wednesday September 19th, 2001 6:06 PM
if you want this to get done, file bug reports in bugzilla.mozilla.org
#56 Re: Should become useful when idle
Wednesday September 19th, 2001 11:08 PM
This is good.
#50 Quick-launch on MacOS/Linux?
Wednesday September 19th, 2001 8:06 PM
Would it be possible to implement Quick-launch on !Windows?
It's a neat feature to have, it shouldn't be reserved to the windows world...
#57 Re: Quick-launch on MacOS/Linux?
Wednesday September 19th, 2001 11:10 PM
I think the traditional Mozilla response is, "program it yourself."
#61 Re: Re: Quick-launch on MacOS/Linux?
Wednesday September 19th, 2001 11:42 PM
No, that's your pat response Tanyel, not ours.
Our response is something like "Mac sort of already does this, when you close the last window on mac apps they stay open and available in finder and the menubar. Linux could do turbo, maybe even better and there are bugs filed to make this happen. If you've got the know-how and time to help implement or help with this or any other requested features we're accepting patches :)"
You can find this and other turbo bugs in the dependency list for the turbo tracking bug <<http://bugzilla.mozilla.org/show_bug.cgi?id=75599>>
#67 Re: Re: Quick-launch on MacOS/Linux?
Thursday September 20th, 2001 12:37 PM
>> I think the traditional Mozilla response is, "program it yourself." <<
More like the traditional open source response.
#71 Re: Re: Re: Quick-launch on MacOS/Linux?
Friday September 21st, 2001 3:06 PM
>>>> I think the traditional Mozilla response is, "program it yourself." <<
>>More like the traditional open source response.
And I soooo prefer the closed source response of 'Sorry, you want the program to do something...?!?!?'
#72 Re: Re: Re: Re: Quick-launch on MacOS/Linux?
Friday September 21st, 2001 11:53 PM
In this case, the closed source alternative has already implemented the requested feature, which is what prompted people to request it in the open source software.
#73 Re: Re: Re: Re: Re: Quick-launch on MacOS/Linux?
Saturday September 22nd, 2001 1:15 AM
the top of this thread was a request for implementation of quicklaunch on mac and linux. so in this case, the open source developers have already implemented the requested feature on one platform, which is what prompted people to request it in the other platforms. the open resource response is something like "it's being worked on or at least thought about by people that are capable of working on it. if you are also capable of working on it and don't want to wait for someone else to do it then feel free to jump in with a patch."
#74 Re: Re: Re: Re: Re: Re: Quick-launch on MacOS/Linu
Saturday September 22nd, 2001 11:34 AM
That seems acceptable to me. I assumed the focus of the discussion changed after "And I soooo prefer the closed source response of 'Sorry, you want the program to do something...?!?!?'"
#75 Re: Re: Re: Re: Re: Re: Re: Quick-launch on MacOS/
Saturday September 22nd, 2001 12:27 PM
So let me get this straight, purely for the purposes of petty victories: Someone says 'I wish we had -turbo for Mac, Linux. You said, incorrectly, that the response would be 'Do it yourself' which you claim is the usual Open Source one. I say 'Better than 'You dare to actually presume to want something out of a product you just use' like the closed source response'. You say 'Actually the closed source version already has the feature' which is totally wrong since both IE and Netscape have -turbo for windows, and neither has it for Mac / Linux (the latter where IE is significantly more functional). And the reason for this is that, er, you interpreted my comment as being about something completely different from the one I was replying to. Methinks it is your comments reinterpreting what people say, and responding to completely different requests than those made.
#87 Re: Re: Re: Re: Re: Re: Re: Re: Quick-launch on MacOS/
Monday October 1st, 2001 9:42 AM
"So let me get this straight, purely for the purposes of petty victories"
I am glad you realize that is your goal.
"Someone says 'I wish we had -turbo for Mac, Linux. You said, incorrectly, that the response would be 'Do it yourself' which you claim is the usual Open Source one."
I said, "I think the traditional Mozilla response is, "program it yourself.'" Do you think I am wrong about what I think? You would have a better chance of getting your "petty victory" if you claimed I was thinking wrongly but you would still fail because I am right.
"I say 'Better than 'You dare to actually presume to want something out of a product you just use' like the closed source response'."
You said, "And I soooo prefer the closed source response of 'Sorry, you want the program to do something...?!?!?'"
"You say 'Actually the closed source version already has the feature' which is totally wrong since both IE and Netscape have -turbo for windows, and neither has it for Mac / Linux (the latter where IE is significantly more functional)."
The closed-source version is Internet Explorer for Windows and the open-source version is an allegedly cross-platform Mozilla. I said, "In this case, the closed source alternative has already implemented the requested feature, which is what prompted people to request it in the open source software." At this point, you have misquoted me twice, misquoted yourself once, and attributed another person's quote to me in an attempt to gain a "petty victory."
"And the reason for this is that, er, you interpreted my comment as being about something completely different from the one I was replying to."
You typed, "And I soooo prefer the closed source response of 'Sorry, you want the program to do something...?!?!?'" The original focus of the thread was implementation of features on Mac and Windows operating systems. Unless Linux is closed-source then when you focus purely on closed-source software, I think it is reasonable to assume the focus changed. I do not understand "er" but I assume it is a pointless interjection as part of your strategy to gain a "petty victory."
"Methinks it is your comments reinterpreting what people say, and responding to completely different requests than those made."
Youthinks that? I think you are another pointless Mozilla flunkie.
#86 Re: Re: Re: Re: Re: Quick-launch on MacOS/Linux?
by SubtleRebel <firstname.lastname@example.org>
Tuesday September 25th, 2001 11:16 PM
What feature are you talking about?
I am not aware of the "closed source" web browser that has implemented a "quick launch" feature for the Mac or Linux platform.
#79 Some proofs that part of IE is preloaded
Sunday September 23rd, 2001 1:17 PM
Macpeep, I did my homework, as I promised. I used some memory monitor utilities to reveal which dlls are used by Internet Explorer.
So, IE 5.5 on my machine (rather minimal installation, no active desktop) uses 35 dlls. 22 of them are already loaded when Windows 98 startup and used by Windows Explorer memory resident part. Of these 22 preloaded files, 19 are also used by the Mozilla browser. The remaining three are : BROWSEUI.DLL (823926 bytes), BROWSELC.DLL (45056 bytes) and SHDOCVW.DLL (1159168 bytes). Total 1.93 MB.
The question is : are presence of the above three dlls significant enough to support the theory that IE "cheats", regarding to startup time ? The answer is yes. It's user interface (BROWSEUI.DLL and BROWSELC.DLL) is indeed preloaded (macpeep, do you remember that I already said that before?) and, while ui isn't the largest part of a complex application, it might be a significant one. IE probably displays it's ui before it's actually ready to retrieve data. So it needs to have its ui preloaded, because first impression is what it counts.
But those IE specific dlls are not the whole part of the story:
1. Mozilla UI is created from zero because it doesn't use Windows system libraries (the opposite would be against Mozilla philosofy, since system libraries are not platform independed).
2. Essential IE data structures (bookmarks for sure, preferences perhaps) are preloaded at system startup. Please, always remember that.
3. Essential system libraries might contain IE specific functions (undocumented, of course). This is only guessing but I would consider very silly for MS to leave this opportunity unused.
That is all I can do. Surely there are professionals that would contribute much more than me on this subject.
#80 Re: Some proofs that part of IE is preloaded
Sunday September 23rd, 2001 1:37 PM
Cool.. I stand corrected for the parts where I was wrong, and my hat is off to you for doing all this research. Very nice indeed! :) I'm very surprised about the UI preloading part, cause I've coded some MFC apps and typically they load very very quickly. It's usually internal data juggling that makes loading slow.. Possibly, the UI part also contains stuff like history, bookmarks etc., which is probably slower to load and process than the actual widgets and windows.
So, how many megabytes are in fact pre-loaded by IE that are not used by Mozilla? Did I understand correctly that it was only 1.93MB? That's surprisingly little..
#81 Re: Re: Some proofs that part of IE is preloaded
Monday September 24th, 2001 12:43 AM
> So, how many megabytes are in fact pre-loaded by IE that are not used by Mozilla? Did I understand correctly that it was only 1.93MB? That's surprisingly little..
Agree, but afaik this is code only, not data. And, unfortunately we 'll probably never learn if other system libraries contain undocumented IE support. But as I said before, perhaps a clever IE startup procedure would give the impression of lightning fast startup time without having anything but the UI loaded. While I don't appreciate MS tactics of bloating its operating system, sometimes their programming techniques are brilliant.
To further explain why Mozilla startup time is so long, I looked at its dll usage. Mozilla uses 129 dlls (!) and this (together with heavy registry reads, according to a known bug) is the primary reason for its dependency on the hard disk subsystem performance. Indeed, Mozilla startup time is greatly affected by hdd speed, pio/udma/scsi mode, disk fragmentation etc. There may be a big bloat here and, unfortunately, only marginal improvements have been done over the last 3 months (most of them unnoticed, except a recent 10% reduction in reload time).
#82 Re: Re: Re: Some proofs that part of IE is preloaded
Monday September 24th, 2001 9:16 AM
Ah, now I see what you said. 1.93MB is the combined size of the 3 DLL's. The MEMORY footprint could be much, much more.. (or even much less, tho that's unlikely).
#83 Re: Re: Re: Re: Some proofs that part of IE is pre
Monday September 24th, 2001 2:58 PM
Actually, I had a lot more DLLs used by both Explorer.exe and IExplore.exe but not by Mozilla.exe. Something like 13 of them. (I know, I should have been more careful and not lost the data.)
#85 Re: Re: Re: Re: Re: Some proofs that part of IE is
Tuesday September 25th, 2001 12:24 AM
> Actually, I had a lot more DLLs used by both Explorer.exe and IExplore.exe but not by Mozilla.exe. Something like 13 of them.
It depends on the features of the IE you had chosen to install and the activation of \"Active Desktop\". My IE installation is almost minimal and I hate Active Desktop :-)