> One would assume that the software was not downloaded automatically. A
> person, somewhere, decided that he/she wanted the software to begin with.
I thought that the whole point of this proposal was that the software
actually was downloaded automatically, without any action or notification of
the owner of the machine -- an extension of the Freenet "nobody knows
exactly what's stored on their node" philosophy to "nobody knows exactly
what's running on their node".
Stefan brought up the difficulty of preventing endless loops or other
resource hogs, and JF's proposed solution was to scan code before running
it. But as the record player stories in "Godel, Escher, Bach" brilliantly
illustrate, that approach cannot succeed.
Your scanner can always be tricked, it's a mathematical fact, unless you
restrict code to the kind of bounded loops you brought up. But if you only
run bounded-loop code, you're shutting out virtually all useful
programs. Anything that acts like a server. Anything that looks like an
application -- imagine if Microsoft Word said, "Sorry, i'm only allowed to
loop one more time before i have to terminate. Oops, there it goes. Bye!"
However, just because the scanner approach can't work, all is not lost.
A task scheduler might work. Let a subprocess run for a set amount of time,
then swap it out. And like Freenet's "cache data closer and closer to the
requesters" policy, such a network could move the process closer and closer
to the person who's running it -- that way, someone who tries to DOS the
network with hundreds of little dumb programs will find those little dumb
programs moving closer and closer, until only the nearest node is dragged to
a halt.
--
Mike Schiraldi
Verisign Applied Research
_______________________________________________
freenet-tech mailing list
[EMAIL PROTECTED]
http://lists.freenetproject.org/mailman/listinfo/tech
