> On Wed, 7 May 1997, Ian Main wrote:
>
> > On Thu, 8 May 1997, Bryn Paul Arnold Jones wrote:
> >
> > > A better thing to do would be to kill off the latest started process
> > > consuming more that x% of the memory. Ie if you start X, and netscape,
> > > and get X with 50%, and netscape with 30%, and x% is 25 then netscape will
> > > be killed (if set at x% was 30+, it would be X again)
> >
> > Is this a kernel thing though? Wouldn't a user space daemon do this job
> > just fine..
>
> Good point. A really easy thing to do would be to toss a message up to
> kerneld (or another daemon listening on the IPC channel) which could then
> take whatever action the admin likes. If one were to attack this problem,
> that would be the more aesthetically proper route.
Yes, and you could use whatever algorythm you like to get the job done --
wouldn't be hard to make it really flexable.. like a list of processess
that should never be killed, a list that should be killed first, and an
algorythm for selecting processes not mentioned in the list. I don't know
why all these people are debating it on the kernel list. If you need such
a thing.. go code it (and make it configurable while your at it.) :)
Ian