* On 2002.04.01, in <[EMAIL PROTECTED]>, * "Rocco Rutte" <[EMAIL PROTECTED]> wrote: > > It may sound funny, but I really saw some Linux guys talking about what > would be necessary to replace a kernel 'on the fly'. Not that it does > make lots of sence or is extraordinary usefull, but to some of them > uptime is all that matters...
NetBSD used to allow you to cat vmunix >/dev/kernel to reload the kernel. It also forced a reboot, too, but I thought it was a neat trick. I don't care about uptime per se, but keeping my processes' state would be valuable to me, personally, and could be really quite nice for servers. It's an interesting lateral approach to the checkpointing problem. (Or maybe it *is* the checkpointing problem... I don't know the details of what you're describing, but it could go either way, I suppose.) -- -D. [EMAIL PROTECTED] NSIT University of Chicago
