jdii1215

I have noticed a few things: If my netscape gets locked as a site tries to feed ads or the site gets hung contacting an ad site, and I kill it with a code 15, I get little of the RAM freed. OTOH, IF I pre-empt it by killing with code 9, I get the RAM back. Similarly, if I have a browser process on top of a mozilla-bin and kill the mozilla-bin, the browser does not get pre-empted-- nor does it run, it needs Mozilla's underlying code to function. So, I use pstree also, these days, when tracing layered calls of processes to see what the base of the tree for the subprocess chain I need to kill is. I have seen X "balloon" over time like this, if an ersatz Java virtual machine and a Sun Java vm fight, but usually it is something like part of a functional process chain got killed and not all of it. Even with mozilla itself, if one kills wrong part of chain one gets only part of the RAM that the whole chain used. This can be true of any process chain. What also happens is that when things get real RAM starved, one needs to preemptively kill the culprits from a console as root, simply because a pure root console login can take priority over a user that is su'd even-- if need be. Do be careful what you kill, though, you can reboot your box accidentally by killing a process that the kernel is treating as a core process.