Skip to main content

Usability issues in Java 6 update 10 build 10

14 replies [Last post]
cowwoc
Offline
Joined: 2003-08-24
Points: 0

First of all I must say I am shocked with the performance of applets in 1.6.0_10-ea-b09.

Between upgrading my computer recently, to the performance boosts of Java6 and update N applets now load *very* fast.

I thought this had something to do with the new applet plugin, so I disabled it. I then disabled Java Quickstart. I then disabled Vista Superfetch. I then deleted both my FireFox cache and applet cache. Finally, I disabled the Quickstart plugin for FireFox.

To my surprise, applets only take 2 seconds on a cold start with all of the above disabled. That's only 1 second worse than Flash (!!) and quite acceptable in my view.

I did, however, run into the following problems:

1) Why is it that http://louisville.edu/~ahdeso01/applets/HowTo.html loads in 2 seconds while http://java.sun.com/applets/jdk/1.4/demo/applets/BarChart/example1.html loads in 4 seconds? I don't believe network latency is the culprit. The latter seems to load a lot more from the hard-drive than the former.

2) Under Vista, Java Quickstart is disabled in the Java Control Panel but not as a FireFox plugin. I am expecting the Windows service and FireFox Plugin to either both be enabled or disabled.

3) Java Kernel seems to install the Java Quickstart service before the full JRE finishes downloading. I suspect that if a user reboots before the download completes Windows will try running a nonexistent executable.

4) Starting an applet immediately after installing Java Kernel seemed to be much slower than starting an applet after installing the full JRE. Why would this be? One would expect the opposite to be true.

5) I didn't like the fact that Java Kernel kept on popping up "downloading components" dialog. I have nothing against the dialog, but it should pop up at most once per applet. The added problem with popping up the dialog multiple times is that its progress bar becomes meaningless (it doesn't reflect the real time I have to wait). Have you considered doing a static bytecode analysis to detect all the dependencies up-front and download them all at once?

6) Finally here is the biggest culprit I found :)

If I try hitting http://java.sun.com/applets/jdk/1.4/demo/applets/DitherTest/example1.html in FireFox without a JRE installed I get told I need to install a plugin. I click on the applet box to install the plugin. I get told "The following plugins are available: Java Runtime Environment". I click on Next. I get told "No plugins were installed. Java Runtime Environment - Not Available [Manual Install]". Clicking on manual install will direct me to http://java.com/en/download/windows_manual.jsp?locale=en&host=java.com&i... -- Now I am expected to download a binary and run it manually then go back to the website and reload the page. Needless to say no sane user will do this.

Contrast this with Flash. I visit a page without flash installed. I get told "The following plugins are available: Adobe Flash Player". I click Next. I get prompted to accept a license agreement directly inside FireFox's plugin installer. I click Agree. "Downloading Adobe Flash Player". A few seconds later "FireFox finished installing the missing plugins: Adobe Flash Player -- Installed". I clicked on Finished. The website reload automatically and the flash applet displays properly. The key point about Flash's installer is that it is integrated directly into FireFox's plugin installer. Why can't Sun do the same with the Java Kernel installer?

I would appreciate your feedback on as many of these as possible before I go ahead and file official bug reports.

Thank you,
Gili

Reply viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
dnp
Offline
Joined: 2006-05-05
Points: 0

In reference to (6):

Our automatic JRE installation for Firefox used to be the same as the what you described for Flash.
Due to security reason, Mozilla.org had recently stopped support one of the XPI mechanism that we used for JRE automatic installation with Firefox (via Firefox' Plugin Finder Service -- PFS).

The behavior you saw is the interim solution while we are rolling out our new approach (in 6u5) and working with Moz.org to get PFS to auto find us. The bug associated with this work is: 6637304.

Note that, Flash is facing similar situation, that is:
- The behavior you prefer is now only applicable for FF2. And may cease to work with FF2 eventually.
- When FF3 tries to find/auto-install the plugin for Flash, it will fail and fall back to manual install, which in turn, direct you to a site to download/install the sw manually.

--
dnp

cowwoc
Offline
Joined: 2003-08-24
Points: 0

dnp,

It isn't clear what the security reason was but I am surprised that the installer would stop working for Java yet keep on working for Flash. Why is that? Are you saying that the old mechanism exposed a vulnerability and Sun chose to disable it immediately until the new installer was ready?

Also, bug 6637304 is marked as fixed so I'm surprised that update 10 build 10 doesn't include it. Are there plans to roll it out in update 10 in the near future so we can test whether this works properly?

Thank you,
Gili

ingridy
Offline
Joined: 2004-11-17
Points: 0

Bug#6637304 has been putback to 6u5 and waiting for java.com group to test it thoroughly. It will be putback to 6u10 after the testing is complete.

Hope it helps,

cowwoc
Offline
Joined: 2003-08-24
Points: 0

Did the putback for Bug#6637304 take place yet? If so, which build of update 10 contains this fix?

trembovetski
Offline
Joined: 2003-12-31
Points: 0

It should be available in 6u10 b14.

cowwoc
Offline
Joined: 2003-08-24
Points: 0

I just re-tried #6 and it seems to be completely broken in FireFox 3.0 beta 5. I visit an applet with no JRE installed, I get prompted to install a plugin but when I click "Next" the installer dialog just shuts down without initializing the JRE install.

I suspect this is a FireFox3 bug though, because when I try installing the Windows Media Player plugin I get the same behavior.

Can anyone confirm?

whisper
Offline
Joined: 2008-09-08
Points: 0

This and several similar issues are being reported with rather alarming frequency in the FF3 forums. (I'm having one of those similar issues myself). I would say the problem is an FF3 issue.

kbr
Offline
Joined: 2003-06-16
Points: 0

Thanks for your feedback. We're discussing the various aspects internally as they relate to the quickstarter, installer, Kernel, and plugin.

A couple of plugin-specific replies:

1) I can't reliably reproduce a difference in startup time between these two applets. Note that I am measuring warm start as opposed to cold start, though starting a fresh IE process each time.

6) Our Firefox auto-install process is a weak link that we are working on. Note that if you just install the JRE on Windows, Firefox will pick it up and use it automatically. The Deployment Toolkit JavaScript addresses this problem to a certain degree. I agree that on Unix platforms the fact that we don't put the appropriate plugin symlink in place upon installation (which is difficult to do in the general case, but an 80% solution is possible) is a weak point. I've filed 6652611 to track this.

cowwoc
Offline
Joined: 2003-08-24
Points: 0

1) Yes, the figures I was referring to were for a cold start.

6) Is there a BugParade issue for the FireFox auto-install process? By the way, all of the above issues were reproduced under Windows Vista (so I didn't run into any of the Linux-specific problems you mentioned).

deployJava.js is a good fall-back in case your particular browser does not have good Java integration but I don't think it is sufficient in its own right. deployJava.js is also great in case you want to measure JRE usage of your website visitors regardless of whether they happen to download your application and I think that has value as well.

sgoethel
Offline
Joined: 2003-07-07
Points: 0

Shall be fixed with http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6668037,
which is submitted for b14.
Measured cold startup times for both applets are different, indeed.
The NewPlugin now consumes less than 200ms for initialization, which is good.
Overall, it took me around 1s for the first mentioned applet, and 2s for the second.
I haven't measured these overall startup times with my guts, so take it with a grain of salt,
i.e. from the click until something is visible.

rogyeu
Offline
Joined: 2006-07-30
Points: 0

http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6668037

Repost the link to bug as the above link has a comma behinds it.

jjburke
Offline
Joined: 2004-03-16
Points: 0

I also had problem #6. Installed everything then had to reinstall the Java Runtime. No big deal though.

Some additional items.
7. The old license screen for the full JDK. Where JRE (I think) has the pretty license screen.
8. It puzzles me fact that every Java update uses INTERNET EXPLORER to display the help or about java web page. FIREFOX is my default browser.

Thanks,
Jim Burke

kbr
Offline
Joined: 2003-06-16
Points: 0

> 8. It puzzles me fact that every Java update uses
> INTERNET EXPLORER to display the help or about java
> web page. FIREFOX is my default browser.

This is 6640156, fixed in 6uN build 11.

qu0ll
Offline
Joined: 2006-12-09
Points: 0

Excellent comments. My thoughts:

1) I guess the second applet requires more classes to be loaded than the first. I haven't checked the source but doing some (simple) graphics must involve more classes than a basic Swing applet.

2) I think Java and Firefox integration is poor. Firefox *runs* Java applets very well once Java is installed but getting it installed and configuring it are somewhat painful exercises. This relates to (6) as well.

3) I don't know much about this one.

4) I haven't noticed this.

5) Totally agree. It is very annoying the way it works at the moment. And it seems to take longer to download the required classes this way than if the entire JRE was downloaded in the first place.

6) Totally agree and this applies almost equally to the IE browser in that the installation of the plug-in is not seemless like it should be. It should be a one click operation... "This site requires Java. Click to install." and that's it... Java gets installed and the applet is displayed without any further user interaction. Until this seemless behaviour is achieved Java will not be taken seriously for RIAs within the browser context.

--
And loving it,

-Q
_________________________________________________
Qu0llSixFour@gmail.com
(Replace the "SixFour" with numbers to email me)