RE: galf once more: application launch feedback

["Ray Strode" <halfline hawaii rr com>]

> To give more examples where this fails, consider people who run
> distributed.net or Seti clients as a low-priority idle background task
> on their computers (not that uncommon). They would always have an
> hourglass mouse pointer this way!
That's not really a good example.  As I said the CPU limit should
be user definable, so if the CPU is always at 5%, then the cap should 
be set higher than 5%.  With a little logic written in, it could probably 
be made to automatically detect a CPU load minimum and automatically raise
the bar.   

> As Seth says; the application launch feedback is to notify you that the
> application you wanted to launch is in fact launching (something that
> you may not know), not that your system is busy (something that you most
> likely already know when you run a distributed.net or Seti client).
Well, that's obvious by the name. :-)  It wouldn't be called application
launch feedback if that weren't true.  I don't agree with you completely
though.  I think if you are launching a program, then you most certainly
atleast expect it to be launching, and either method would provide adaquete
verification for that case.  My method would also provide feedback if there 
is some task going on that is busy.  As Seth suggested it could be an hourglass 
for launching and some sort of hybrid pointer for things that happen while the
program is running.  

I think that maybe what would be really cool is if it had a busy cursor for 
the programs that are busy, but if you iconify them or if they lose the focus, 
then your cursor turns back to complete normal and the tasklist changes the app
icon to a busy signal.  Then when it's done, the tasklist could blink or something
and restore the app icon.  

I think this could be achieved by looking at a processes CPU load, but I'm 
wondering, how much overhead would be caused by doing such monitoring?  Who wrote
libgtop?  Could it be used for this, and would such a technique be efficient? 

--Ray