From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jim Choate To: <9fans@cse.psu.edu> Cc: Subject: Re: [9fans] 'wall' messages In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Date: Wed, 8 Oct 2003 17:59:32 -0500 Topicbox-Message-UUID: 685c2ea6-eacc-11e9-9e20-41e7f4b1d025 On Wed, 8 Oct 2003, Charles Forsyth wrote: > you might think to notify people that the cpu server is going down > by having it access devices imported from the terminal, but > don't forget that the file server has no such access, and in many > ways it's far more serious to have the file server go away. > most reasonable solutions to the latter will deal with the former, > including some of those suggested so far. I have to agree with this. The main reason that I find this sort of fail-over critical is if a process does die, if the file server is still around we have some hope of restarting the job on another set of cpu servers without the user even being aware, other than longer execution times. If the file server dies the namespace is gone so we've no way to ensure the intended namespace, we may not be able to build it up again without user intervention. I find this most critical within the context of what I call 'free run' agents. I envision a point when there will be enough 'public' servers available that processes will be able to run forever (for all intents and purposes), this is most important for example with anonymous remailers, black nets/data havens, etc. In addition it would allow the creation of 'public' search engines that simply were not censorable because they weren't running in any particular place [1]. This of course means there needs to be a mechanism to track restarts and alert the user if they're having too many. [1] My personal belief is that this will be how the first self-aware AI comes about, a pure accident due to programs interacting and trying to survive. We'll get up one morning and all the phones will be ringing ;) -- -- God exists because mathematics is consistent, and the Devil exist because we can't prove it. Andre Weil, in H. Eves, Mathematical Circles Adieu ravage@ssz.com jchoate@open-forge.com www.ssz.com www.open-forge.com