[MUD-Dev] Re: cellular automata as universe models

[Joel Kelso]

James Wilson wrote:

> Hi all, I've been thinking about how one might use cellular
> automata to model a universe. By 'cellular automata', I mean a
> set of nodes where a given node's state changes are completely
> determined by a set of neighboring nodes. (The research community
> uses the term a little more carefully than I do. Purists, please forgive
> my insolence.)

Have you read "Permutation City" by Greg Egan ?

<snipped design bits for cellular-automata based MUD>

> As a 'realism' constraint, there is a real-world parallel; namely, information
> and thus causality cannot propagate faster than the speed of light, so there
> is no instantaneous causality across remote areas of space. (Quantum
> correspondences don't let you out of this either, unfortunately, but that's
> highly OT. *heh*) Perhaps one could loosen the correspondence between
> spatial locations and nodes - allowing mobile nodes, wormholes, dynamic
> changes to the node graph, and the like - without breaking the underlying
> synchronization model.

If you disallow non-local access, you might get another advantage: the abilityto
simply partition the processesing for your world onto multiple processors.
Inter-process communication occurs only at well-defined boundaries (although
you could shift the boundaries dynamically to do load balancing), and by
a well-defined protocols.  Just a thought.

> To be perverse: at one extreme, each object could be a single mobile node,
> but this would obviate any gains realized by serializing groups of spatially
> local objects, and seriously complicate the process by which events inspect
> their environment.. At the other extreme, nodes could be so large that
> serialization of their state changes would be tantamount to serializing the
> whole db, which of course is an option but one I am interested in avoiding.
> The useful case is where nodes are large enough to conveniently serialize
> db state changes while not terribly inconveniencing script writers.
>
> The node and its neighbors would have to be understood as the 'environment'
> available to an event, and script writers would need to work under the
> assumption that information from outside this local environment is unavailable.
> This is essentially the same constraint that human users of VR environments
> (and real world systems) operate under; that is, one's responses are determined
> by one's internal logic (mind, instincts, physical laws, etc) and one's
> immediate environment (through sensory data or physical interactions). I'm not
> sure how this constraint could translate into a scripting system, however. One
> would need to remove direct references and proxy all object accesses through
> the environment; how could this be done efficiently?

I have to admit that when I first read you post I thought: "what do CA have to
dowith MUDs ?" ... but the possible serialisation and parallelisation advantages
of a cellular world with no non-local references really make this worth thinking
about.

Joel Kelso

-- joel at ee.uwa.edu.au --------------------------------------
"... great Scott, he's turned into _more than one person_ !"
"Well, there was always enough of him."
 - the Goon Show
-- http://ciips.ee.uwa.edu.au/~joel ------------------------