Yup, watching via procexp, you can see clearly that we launch iexplore.exe, he spawns ieuser.exe, and then iexplore.exe dies; ieuser.exe then launches another iexplore.exe and that's the one you see on the screen.

http://msdn2.microsoft.com/en-us/library/bb250462.aspx explains ieuser.exe and Vista's protected mode.

I think the best solution here is to use jobs, as described here http://benjamin.smedbergs.us/blog/2006-12-11/killableprocesspy/
It's based on the last few paragraphs of this documentation: http://www.microsoft.com/msj/0698/win320698.aspx

The easiest thing might be to take the code in killableprocess (or something like it) and compile it into an exe that we can ship in the jar. The Best(tm) thing might be to use JNI, especially if we could figure out how to solve the problem of how to package DLLs in JARs and delete them when we're done.

I discuss the idea of JNI in more detail in SRC-22.

Fixed in rc revision 1984, which unfortunately also created SRC-343. But that's considerably less important, IMO.

Jean-Charles Meyrignac is reporting that this is still broken, so I did some more work to make it more reliable (by fixing SRC-343) in RC revision 2001.

I'm having issues with this on Vista. The manual test cases in killableprocess.cpp doesn't work.
Opening a cmd prompt and just typeing
C:\apps\kp.exe c:\progra~1\intern~1\iexplore -new http://www.google.com
will launch the browser but detach right away. Most of the time anyway. As I'm entering this comment I got "the other" out of two errors I've been seeing:
QueryInformationJobObject failed with error 234
And it just dies. Sometimes I can see the browser window opening but it gets killed before anything but the frame of the window is painted to the screen.
I'm not sure this issue is completely solved yet.

According to the last comment this issue is not completely fixed. As I don't have Vista could you please check Dan what's the problem?