This is a good point as far as it goes, but it needs to be more radical.
Scale inherently changes problems. A VRMUD will be of such a large scale
that i doubt even client side processing will be enough. Then there's the
need to proxy devices not capable of the client side processing. This
leads to an event more distributed architecture where agents are put
anywhere and in any number. They need to support protocols to establish
roles, replication, load balancing, partitioning, hand offs, mobility,
etc.