Evolution is a process that explores the possibilities inherent in the medium in which it is embedded. Evolving populations of replicators constantly explore variations around their current forms, without the limitations of preconceptions. When embedded in the medium of carbon chemistry, evolution ``sees'' the physics of the natural material universe: the laws of chemistry, such as that carbon forms four single bonds in a tetrahedral configuration; the laws of thermodynamics, such as that entropy spontaneously increases, etc.
However, when embedded in the medium of digital computation, evolution ``sees'' a completely different universe with different laws. There are no laws of thermodynamics. There is no material on which to base a chemistry. It is a logical informational universe rather than a material universe. The ``physics'' that evolution experiences when embedded in the computer consists of the logic implemented by the processor, the unique non-Euclidean topology of the memory space, the rules for resource allocation embedded in the operating system, time based on the CPU clock cycle, etc.
When embedded in the digital medium, evolution does not experience the material of which the computer is constructed. The computer could be built out of large scale integrated circuits, transistors, vacuum tubes, or mechanical switches. As long as the hardware implements the same logic, it will be indistinguishable to evolution.
Seen from our point of view, the different physical technologies would at least differ in their speed. But seen from the point of view of evolution, the unit of time is the CPU clock cycle. Whether the CPU clock cycle takes a nanosecond or a minute, experienced from within the computer it is the same. However, in the case of a network of computers, relative CPU clock speeds do become important.
The evolving digital organisms occupy the space of the RAM memory of the computer. Because we assign sequential numerical addresses to locations in this memory, people tend to think of it as a one-dimensional Euclidean space. However, it is not.
The topology of a space can be understood in part by examining the distance relationships between sets of points. In computer memory space, there is no meaningful concept of linear distance. The most appropriate analog of distance is the time that it takes to move information between points. Thus time becomes the metric for distance in memory space.
In the memory of most kinds of computers (with ``flat'' memory), all pairs of points are equidistant, regardless of their actual locations in memory. In a computer with segmented memory, such as the Intel 80X86 processors, there are two distances between points, depending both on the relative actual positions of the points and the frame of reference. This shows us that the space is clearly not Euclidean.
If we consider the topology of the memory space of the global network of computers, ``cyberspace'' at large, it is very complex and dynamic. The time that it takes to transfer data between the memory of two computers will depend on their relative physical locations on the net, and the traffic conditions at the time of the transfer. Transfers within local area nets will generally be faster than transfers between more distant locations, but both will also depend on network loads at the time.