[haiku-development] Re: On timeslices and cycles
- From: Danny Robson <danny@xxxxxxxxxxx>
- To: haiku-development@xxxxxxxxxxxxx
- Date: Mon, 16 Mar 2009 12:20:57 +1100
On Sun, 15 Mar 2009 21:08:34 -0400
Ryan Leavengood <leavengood@xxxxxxxxx> wrote:
> On Sun, Mar 15, 2009 at 8:10 PM, Danny Robson <danny@xxxxxxxxxxx>
> wrote:
> >
> > Is it worthwhile neglecting a feature because people will abuse it?
>
> [..] We already have a big list of features
> that everyone wants but no one has yet implemented, so I don't think
> something like this would be high on the priority list.
> [..]
> So I don't think the cost/benefit ratio is there, based on what I've
> read in this thread.
Sorry, I should probably have prefaced this with a note that I'm not
suggesting it's critical, that someone must implement it, or that it's
more important than anything scheduled for R1. I just think it can be
very useful for some (minority of) developers.
> > Any issues arising from misuse of these features essentially become
> > the application developer's problem.
>
> That kind of logic can be a cop-out sometimes. You could also blame
> developers for misusing memory and CPU before we had memory protection
> and preemptive scheduling: "it is the developers problem to ensure
> they don't clobber other program's memory or hog all the CPU in
> cooperative multitasking." If that had worked out so well the newer
> technologies and operating systems would not have developed. I hope
> you agree that restricting developers a bit in those areas is good.
No memory protection and cooperative scheduling have the potential to
make a system dead or unusable. With hard affinity the only entity
adversely affected should be the one using the feature.
- Danny
Other related posts:
