| |||||||||||||
KHB: Failure-oblivious computingKHB: Failure-oblivious computingPosted Jun 29, 2006 6:56 UTC (Thu) by nix (subscriber, #2304)Parent article: KHB: Failure-oblivious computing
`A little mind-boggling'? I'll say. It sounds like a very good idea for a large class of applications, and for a large subset of things within even applications which might otherwise not seem to be candidates.
Thanks for drawing my attention to this, even if it *is* breaking my brain.
(Log in to post comments)
KHB: Failure-oblivious computing Posted Jun 29, 2006 10:57 UTC (Thu) by kleptog (subscriber, #1183) [Link] I agree, it seems just wrong somehow.
OTOH, I can see some useful applications. Imagine if in addition to hiding the error, it logged it somewhere. If the overhead isn't too great, you could just enable this on your server and scan for failure reports daily.
Most of the time people disable core dumps from broken programs because they take up too much space and they need processing to be useful. If just a small amount of specific logging were included, some simple tools could watch for repeated failures and even automatically generate bug reports. All without interrupting service. It almost seems the first step to self-fixing computers.
My brain still hurts though, just thinking about it.
|
|||||||||||||