The architecture of cyberspace can change, but not indefinitely. I argue that some of the regulations by architecture have limits due to structure of code and the tradition of sharing among coders.
The amount of control has a limit that is the computing resources available for performing it. Computer Scientists analyze the performance of systems using various methods; they are concerned with the time or space complexity of the system. Crudely, that is how much time and memory a code takes. When the system is distributed or parallel things get even more complicated because a new resource which is the network itself must be considered; furthermore concurrency and communication among processing units must be taken into account. Internet is such a system. The more complex tasks performed on Internet, the more it resembles a general purpose distributed system.13 My claim is that the complexity of control cannot be arbitrarily increased, as is the case with any algorithm. I will not offer an exhaustive argument here, but I would like to be a bit persuasive. The simple explanation is that there are only this many resources, and using them for control will exhaust these resources beyond a certain point thus preventing a more accurate control. This I think is a mathematical limit to regulation by code.
As an intuitive support to this idea consider the following. A great part of your thought has to be sub-conscious because you, as a computing device, cannot afford having conscious control over every computation in your mind. Some of the things you think, you have no idea that you are thinking them, because it would require a much larger brain if you had the ability to control those lower level features of your thought. And you would probably run a lot slower. There is a limit to how large your brain can be, and how slow you can think because you need to survive. All you are conscious of is a vague description of your state of mind. You cannot have perfect control over yourself, and the Internet cannot implement perfect control.
Another aspect of distributed systems make it even more difficult to perfect control. The Internet is a distributed system, because it cannot be realized as a single computer, a UNIVAC that every other person logs into. That is impossible because a single machine has too little resources. The collective resources of a network are much bigger; that is the whole promise of distributed computing. However, in order to make it work at all the system must have some basic protocols that will interoperate. That is what Internet is like. A high level process can work on the system to the extent that it is permitted to run on individual nodes of the system. If a node declines to run the process, it cannot be controlled by that process. This is the sovereignty of individual machines. Unless there is a law enforcing you to provide a backdoor to a government agency, you do not have to give the outside world any control.14 This means that if a set of machines do not abide by a process of control, then their operation is to a large extent uncontrollable. Of course, one needs control over his software for this. This is the limit due to openness of the distributed system.
An essential component of cyberspace, the people, strengthen these limits. Code writing itself cannot be strictly controlled because there are many individuals writing code. That is also Lessig's argument. My contribution is that at any moment the coders can find a way to construct a new network with the desired freedoms and share that knowledge. Of course, the basic protocols should be operational, otherwise the fellow hackers cannot connect. Lessig says that only hackers could circumvent copy protection and thus majority of the population would effectively be regulated but this is not the case. Coders have a tradition of sharing their code with the non-coders. Ordinary users usually benefit from such circumvention. In my country almost every PC with a TV tuner has the software that can decode encrypted TV channels. It is not because it is full of hackers; it is for there is great demand for such code. It seems to me only as a continuation of the resistance among coders against codes that impose undesirable regulation; such as overpriced software with copy protection.
There is another happy obstacle, to copyright regulation by
code. Stefik's horrible trusted systems can never be fully realized
because copying is a basic operation that the owner of any
computer can execute. There is no way to prevent copying data if it is
in memory. Thus, trusted systems cannot be implemented without replacing
every bus and processor in the world and then making each copy
operation a database transaction over the network which is impossible
because computers would be 
times slower.15