Possible JSF bug: link failure in Internet Explorer

Question

Greetings:
I have a form which, among other things, contains a table (generated using the <h:datatable> tag). One of the columns in this table contains Command Link (h:commandLink> whose action attribute points to a method in one of my Beans. The command link contains an onmouseup event handler which displays a dialog box for confirmation before actually going to the link.
When I click on the lcommand link in any Mozilla- based browser (Mozilla, Firefox, etc), the confirmation box comes up and, when ithe user clicks on it tmy Bean's method is called. This behavior is expected and works well.
Unfortunately, in Internet Explorer, the confirmation box comes up, but my Bean's method is never called! This is the first time I have seen a custom tag of any kind (in this case, <h:commandLink>!) work in one browser and fail in another.
Could someone tell me why this is or (preferably) how to make my commandLink actions work in Internet Explorer???
Thank s anyone who can help.
Factor Three

Answer 1

Actually, I found the problem. It is a combination of poor implementation of the <h:commandLink> tag's Javascript and a difference in Browser operability.
The Browser difference is apparently in the way different browsers handle events. In Mozilla- based browsers, when you have an onmouseup event and an onclick event, apparently both event handlers get called. In Microsoft- based browsers, however, apparently if you have an onmouseup event and an onclick event only one of the event handlers (in this case the onmouseup event) gets called. Why Internet Explorer's developers made it behave this way is a mystery, because it goes against the way most Windows applications handle such events (usually, they process any mousedown events that exist, then any mouseup events, then mouseclick events!).
The result is that in a Mozilla browser, my onmouseup event handler gets called, I click on my Javascript- based confirmation box, then the browser activates the onclick event handler that the <h:commandLink> tag generates. In Internet Explorer, however, only the onmouseup event handler gets called; since <h:commandLink> places the code for submitting my form in its generated onclick event, that code never gets called and Internet Explorer never calls my Bean method.
The poor implementation of <h:commandLink>'s Javascript comes from the fact that this particular tag does not allow the declaration of onclick events in its TLD. Why the creator of this tag did this is a mystery to me, because all other tags in this category allow such event declarations in their TLDs. It could be said that because <h:commandLink> provides its own onclick event it is logical not to allow such events. This logic is flawed, since other link tags (like <h:commandButton>) also provide their own onclick events, yet they allow programmers to define their own onclick events. I have looked at the generated event handlers for these tags; all they do is take the event handler you define, append a semicolon at the end, then append their standard event handling code behind yours. For some reason, the <h:commandLink> tag does not do this.
The reason <h:commandLink>'s Javascript handling is poor is because it prevents the addition of any kind of client- side validation, confirmation, or any other operations someone may wish to do using this tag prior to invoking any server- side functionality. The reason why I put in the onmouseup event handler in the first place was because in my user interface I needed to display a warning to a user and to get confirmation that the user really wanted to perform a certain operation before actually going to the server and performing it. Since <h:commandLink> is the one tag that invokes a required bean method and passes it parameters based on an item's position in a list, I had to use that tag and, since it didn't allow onclick event declarations, I had no choice but to use onmouseup.
If <h:commandLink> had been better implemented (in other words, if it allowed the declaration of onclick events like other tags do) then I would not have been left open to the problems associated with the different browsers' event handling. For this reason, I can say that I have, indeed, found a JSF bug -- though it is a rather subtle one.
In the case of my application, I was able to come up with a kludge which makes it work on all browsers. Basically, I took the onclick event handler code generated by the <h:commandLink> tag and copied it into my onmouseup event handler. Now the proper submits are being done when any browser is used.
Of course, it would be better if someone at Sun were to make a slight change in <h:commandLink>'s implementation, or if Microsoft were to make its browser implementation more consistant with Javascript standards. But then, life for all of us would be better if we had World Peace...
-Factor Three