The Net in Cyberpunk is not particularly well described and is both illogical and inconsistent in its structure. This is odd because, say what you will, the Net and netrunners are perhaps the most cyberpunk aspects of the game. What follows here is my attempt to understand and make the Net useful, at least as I see it. Feel free to disagree with me if you want. I admit I am no computer expert.
So what exactly separates netrunners and the general run of the mill computer user in 2020. From the main rules this is not actually easy to tell. Everyone with computer skills seems to be a netrunner. This should be the furthest from the truth. In fact netrunners will be a minority among computer users as hackers are today.
Netrunners are people who use their knowledge of computers and programming to break into other people's (in this case Corporations for the most part) computers to perform software sabotage or computer espionage. This requires a lot of expensive software and some fairly major modifications to commonly available hardware. A General User is the other 99.9999% of computer users who are on the Net for legitimate reasons.
General users need most of the same hardware systems used by netrunners to access the Net. Oh, sure they can use keyboards to perform the most rudimentary tasks, but a full virtual interface with all its intuitive control is by far the more preferred method. As a result most general users will also own cybermodems, but very different cybermodems than the netrunners. Ever wonder why companies like Ono-Sendai are allowed to produce cybermodems, especially if they will all be used to break into the data banks of other corporations. Most police forces, not to mention Interpol, would very quick to close such a company down. (Sure one could simply say "Hey how could we to know what they would be used for?" But like a company who produce high powered assault rifles or illegal drugs the producers of cybermodems still must shoulder some responsibility for the use of their products. And even if the law couldn't enforce this the other corporations surely would.) Ono-Sendai is allowed to produce its product mainly because an off the shelf cybermodem is virtually useless to a netrunner.
But why? The off the shelf cybermodem is designed to be a peripheral to a computer. (Come on admit it, cybermodems as described in the rules are pathetically inadequate for any real computer tasks. They are able to perform a very limited range of functions with a high degree of effectiveness, selectivity breeds strength but not flexibility.) Cybermodems have limited computing power and very limited memory capacity. The netrunner's version is designed to do a single job with the maximum amount of portability. (Who knows when Netwatch will break down the door, or when you will be required to break into a site to get on a dedicated terminal.) Therefore what good is the commercial model. It allows a computer user to access the Net with full virtual reality interface, and with cyberpunk's direct neural input all the senses can be used. The general user makes use of his far more efficient computer to manipulate data and for greater storage capacity, while linking through the cybermodem to allow faster control. This is also much closer to William Gibson's version of netrunners. (Case has a Hitachi computer directly wired to his cybermodem for the more mundane tasks of information retrieval.)
Okay, but why can't the shelf model be used for netrunning? All commercial computer equipment is designed to identify itself when authorized personnel request it. This is not so far fetched. Modems communicate with each other today, and I was recently told that some newer software is able to identify Intel chips (so that the program will not work for competing chips). And look at the dreaded Win95 and 98, it can identify hardware systems, even on older models that don't have the plug and pray adapter. And now we have Intel's masterpiece, the Pentium III processor, which can not only identify itself but let anyone and their mother know what software you are running. So here we have commercial hardware that basically screams its ownership and operational capacity to anyone with the right to ask. So who can ask? Netwatch and the other online police forces, any local and international police forces and any corporation willing to pay for the right. This means that any idiot who tries to run the Arasaka datafortress with a store bought unit (probably already registered at purchase) will soon be visited by kevlar armored stormtroopers with very big guns.
This is why a netrunner's cybermodem requires a great deal of modification. First a custom chip set is needed to increase the speed (assume all commercial models are 2 points slower than listed in the cyberpunk equipment list, with half the wall value). The modem will then need expanded memory capacity. After all as a peripheral the modem only has buffers, not real memory. Finally all the identification chips must be altered. Not removed, altered. Why altered? Well any cybermodem that is inspected will be tested, and if it cannot identify itself in some fashion, it will be declared illegal and the owner charged with a computer crimes violation (regardless of true guilt, this is after all a police state). The modem will also need a new OS. Commercial users do not need all the hacking commands and therefore will not have them. Finally the netrunner's cybermodem will require a new type of routing controller. In 2020 all net time costs money so commercial products will be designed to route directly to the source by the shortest possible course. This is of course very easy to trace. Therefore a netrunner needs to be able to control the path he takes to his target. This allows the creation of decoy routes and makes it harder to trace and a reconnection must be made at each new interchange (more or less as described in the main rule book).
Finally we have software. Commercial users have no need for the more aggressive programs, no kindergarten student should need a jackhammer. Commercial modems are only equipped with defensive and anti-virus software. This of course leaves them naked when compared to the better-equipped netrunners. But any general user who finds he has stumbled into a restricted datafortress will be shown no mercy (although an inquiry will be held, as opposed to the general outcome of a netrunner who dies by ICE, actually netrunners get no trial in most cases).
ICE. The dreaded scourge of a netrunner's existence. But how exactly does the blasted stuff work? Most references in the rules mention surges of electricity down the modem line. But what about fiber optics? Most commercial phone-lines and nearly all server connections can be assumed to use fiber optics (after all the trend has already begun), especially if the population of North America continues to grow. So now you see the problem. How can an electrical charge be directed along a fiber optic pathway? The simple answer is that it can't.
But that cancels out all ICE? Or does it? What if we re-examine the way ICE works. First we will look at ICE and the Human brain, and then ICE and computer hardware.
ICE affects the brain in a simple fashion. It tricks the cybermodem to send an anomalous command to the neural processor (be it the one implanted in the netrunner or the one in the box attached to the 'trodes). This signal sets up conflicting neural activity in the brain leading to short circuits. For those who haven't kept up, a short circuit in the human brain usually leads to things like epileptic fits. Not all that hard to do, even with a simple television today. Certain frequencies and color patterns can cause epileptic seizures in a small portion of the population. With a direct neural connection it becomes that much easier. Make the seizure strong enough and you can inflict permanent brain damage or even induce a coma. But it gets worse. The neural processor is designed to translate artificial stimuli into true nerve impulses, and it is often attached to some portion of the upper central nervous system (CNS). As a result the processor has probably got some form of connection to just about every portion of the human body. Mix a few signals and the autonomic nerve system (ANS) could collapse (think terminal heart attacks). Even though the ANS and the CNS are separate it is possible to exert limited voluntary control over ANS functions, try holding your breath sometimes. It is even possible to affect heart rhythms through biofeedback and emotional states can cause psychosomatic disorders. How much more effective would this be to someone who can record neural activity in test subjects until the correct messages can be found, and who has direct access to the CNS and the brain. If you want to get even more fantastic consider the effects of direct neural manipulation on systems like the adrenal glands or the thermoregulator system. Over stimulating the adrenal glands will lead to borderline psychosis and nausea as the body tries to fight its own chemically induced fear (it should be noted that the effects of adrenaline on human behavior are not as cut and dried as adrenal boosters would seem, flight is just as powerful an urge as fight and with enough adrenaline in the system it will require a cool check to remain in control). If the thermoregulatory systems are affected the netrunner could be in for things like lethal fevers as his body tries to fight off infections that don't exist. It should be fairly obvious from this that 'runners not using 'trodes or a direct input plug are largely immune to ICE. Goggles will have some effect but only at half strength and effect, and keyboard operators are completely immune.
But what about ICE and hardware? Hardware is just as at risk to ICE as the brain and for very similar reasons. Since most modems will use fiber optics it will be impossible to simply burn out a modem and the human using it. Although it might be possible to pay Ma Bell to direct a true laser beam along a fiber optic connection effectively melting at least some of the modem. So how does ICE affect the hardware? Simple, most modern systems use software controllable switches (consider your computer's reboot alt+ctrl+del). By tricking the modems control software it is possible to mess with its functioning. Turn off a cooling fan and burn out a chip, flip the on and off switch rapidly to burn out other circuits. Set disc readers to vibrate to throw heads out of alignment. Set chip based memory systems to rapidly write and rewrite sectors to burn them out. The next method is to affect the modem's software. Introduce viruses to destroy valuable programs or corrupt important data. Add recognition codes with hidden programs to alert any police systems. Any of this can be used to stop a netrunner. And to make things even more annoying simply turning the system off will not solve the problem unless a safe method of access can be found (Most cybermodems will require this feature to be custom installed for precisely this reason).
IG Transformations. Oh boy, what a bad idea. As I understand it the IG transformations are supposed to have set the Net up in a way that mimics the actual spatial coordinates of each computer system. For what reason I am not sure. Let's just consider this for a second. In the Net as described by the Cyberpunk rule book such an idea makes some sense. All a user needs is a map of the city or a general familiarity of its layout and he can find just about anything. All well and good if a city the size of Night City has only 20 or so permanent net locations. This is of course completely insane.
But why? Lets look at today's net (admittedly based upon my limited knowledge). At one end we have the users, sitting at home or at the office with a modem attached to their computer. Some of those users will actually be attached to a network (a group of directly linked computers attached to a modem or directly to a server). The modems will generally be attached to a server via the phone lines. A server is a dedicated computer whose main function is to allow users to hook up to other servers. It is this net of servers located all across the world that forms the Internet. Admittedly this is very simplistic. Hooked into this system are all of the larger computers and at each server we have web pages (sites like this one created by users, but which will be constructed in the virtual language we call the Ihara-Grubb transformations). The servers are all connected more or less directly through the phone companies, but generally using faster communication lines.
So how does this affect 2020. Well the IG transformations work if there are only Datafortresses (a cross between a company home page and a direct access line to its networks). But there are not. Even worse the bundles of fiber optic lines that link this whole mess do not necessarily form road map like patterns. Linked to the Net are all the servers being used by all the users, and of course all their computers. The Net is so pervasive in 2020 that all middle class and higher dwellings will have some form of local server (for example all the houses on a court in the suburbs will be on that court's server, and all of the apartments in an apartment building will be linked to a server). Not only do computers hook up to these servers but so to do phones and televisions. As a result from inside the net an apartment block will look like a Christmas tree. As users log on and off their personal computer representations in the Net will flicker on and off, the building itself will look like a mess of bright wires as the IG transformations try desperately to map them out accurately.
The next problem is security. What if you have a secret facility out in the desert, obviously you don't want anyone to know that it is there. But the net will show that something is out in the desert and it should be possible to use netrunner software to get precise coordinates. And what company in its right mind would put its datafortresses out in the open where any netrunner could find them. More realistically the datafortress the world sees should be a front (not to mention the fact that it can be dropped off the net with the flip of a switch). The fortress should in all likelihood be a trap, or at least the defenses around a private LDL. Even the concept of LDL's makes no sense (see Communication and the Net).
There is an even worse problem listed in one of the rulebooks. A submarine off the coast surfaces and accesses the Net. Once its computers are hooked up (via cellular modems) the sub becomes a site on the Net that can be traced back to its real world location, not very smart if you have a missile sub that has to remain hidden. And what about cellular modem users, they flicker on and off the Net like ghosts. Better yet suppose your netrunner sneaks into a corporate headquarters to access the system directly. Any Corp sysop will be able to look at the datafortress Net representation and say "Holy Shit he's in the building, and he's on floor 10...". This sucks majorly. It's not very practical.
The way I see the net is more like Reboot. You know, that CGI TV show! In Reboot each computer is an island unto itself. Each server and each large-scale (read corporation level) computer will have a representation on the Net. By accessing one of those systems you are able to see access the web of users and lesser systems attached to that system. The IG transformations should not be all encompassing. If a corporation chooses not to register with the system then it will not show up on the Net for the casual user. Therefore cities like Night City are like small galaxies of light, each server staying online almost all the time. But the system has increasing levels of detail. At the lowest level of detail the user gets a map direct from the IG corporation (like Internics today) that shows ever major cluster of systems in the world (forget LEO for now it has far to many conflicts with the cyberpunk view of netrunning to be allowed as is). The next level of detail brings the user a view of each country and then down to states and then down to cities. Inside each city only the major systems will be initially visible but as one sinks into the local grid more and more lesser lights will begin to glow. Drop into a light and access a server, then the individual systems attached to it become visible.
This is a problem. No one explains how the IG transformations actually work. We are simply told that they are responsible for setting the general geography of the Net and they make sure we all see the same thing when we login. But how?
Okay the way I see it is this. IG refers to a new kind of generalized virtual reality language. It defines all the three-dimensional and animation commands, and since it has become the standard everyone sees the same things. But why do netrunners still use this stuff when they design their own software? The main reason has to do with the shape of advanced software design. As computers and computer languages get more complicated the poor pathetic meat human at the other end will require a great deal of automation to help him out. Almost all programming is actually done with component blocks that can be shifted around and tweaked. The blocks are usually assembled by an editor program or an expert system. The reason not everyone makes their own programs (although some utilities and most basic software could be built this way at home) is because building a new module requires some real skill. After all blocks are easily recognizable, tradable and can be reverse engineered. Therefore attack and defense software using this format will be useless very quickly. This is another reason why AI's are so useful, they think in this language and are able to create new modules with ease.
But why are the netrunners using IG programs? The reason is this. Unless a really intelligent programmer comes along and finds a way past the IG system the netrunner's programs will be given random icons to represent them. Pardon? Okay its one thing to write an invisible program, you just assign that program an icon that is computer transparent. IG knows the program is there, that's why all those identification programs are able to find invisible programs. The program is only invisible to the human using the system, not the computer (of course there should also be automated ID programs in all computers but that's another story). So you write a program that uses IG language but you don't assign it an icon, well the other person's system just assigns you one. It could a block or something else as lame. This is why netrunners spend so much time building neat icons. If you have to use the bloody things then you might as well look cool doing it.
But what is this way past IG? Well a truly gifted programmer should be able to find a more direct way to talk to a computer (this is a very difficult task and often nearly impossible so try doubling or tripling difficulty to create). If you could get a program to work that didn't use the IG language (sort of like using your PC to log on to a military computer that uses its own unique programming language and screwing it up, it can be done but I have no idea how), then the IG recognition system in the other user's computer couldn't find or see the program.
I should also note that there are advantages to running you programs without built in icons. They take up half as much space and load up twice as fast (+2 for our purposes). Your reputation will suffer, but you are packing far more firepower.