The Great Firewall's Origins: Documented State Control Versus Rumor

In September 1987, from a computer at a research institute in Beijing, a group of Chinese and German scientists sent the first email out of the People’s Republic of China. The message, drafted in English and German, read: “Across the Great Wall we can reach every corner in the world.” It was a line of pure optimism, the sort of thing engineers wrote when a network first blinked to life. Within a little over a decade the same government whose institute sent that message would begin building, in careful and documented stages, a system whose entire purpose was to make sure that reaching every corner of the world worked in one direction only. The wall in the sentence had been a metaphor for connection. The wall that got built was literal, and it was designed to keep things out.

The system that was really constructed

The temptation with Chinese internet control is to imagine it as a single monolithic black box, sealed and unknowable. It is not unknowable. A great deal of the “Great Firewall” is a matter of public record — laid out in Chinese law and regulation, described in the marketing materials of the Western companies that helped build it, and mapped in fine detail by more than two decades of academic measurement. The censorship is real and heavy; it is also, in its architecture, remarkably well documented.

The formal beginning was a project with a bureaucratic name: the Golden Shield Project, approved by the Ministry of Public Security and launched around 1998, with a trial phase running into the early 2000s. Golden Shield was broader than web censorship — it was a national programme to computerise and network the police, tying together identity records, surveillance and communications monitoring. The part of it that came to be called the Great Firewall, a nickname coined by a Western journalist around 1997, was the outward-facing filter: the machinery that sits at the handful of chokepoints where China’s internet connects to the rest of the world and inspects the traffic passing through.

How it works is not secret, because researchers have probed it from the outside for twenty years. It filters at several layers. It poisons the Domain Name System, so that a request for a blocked site returns a false or dead answer. It blocks specific IP addresses outright. It performs “deep packet inspection,” reading the content of traffic for forbidden keywords and, when it finds them, injecting reset packets that tear the connection down — which is why a page can begin to load and then die mid-sentence. More recently it learned to detect and throttle the very tools built to evade it, fingerprinting the encrypted signatures of virtual private networks and circumvention software and choking them. The list of what it blocks is long and familiar: Google, Facebook, Twitter, YouTube, Wikipedia at times, foreign news outlets, and any page touching the subjects the state treats as radioactive — the events of June 1989 in Tiananmen Square chief among them.

The Western role in this is part of the record too, and it is not a comfortable part. In the early 2000s, as the infrastructure was being built out, American networking firms sold China much of the hardware on which it runs. Cisco Systems, in particular, faced years of litigation and congressional scrutiny over routing equipment and, according to internal documents surfaced in court, marketing materials that referenced the state’s goals of controlling dissent. The companies argued they sold general-purpose equipment; critics argued they had knowingly furnished the tools of a censorship state. Yahoo, separately, handed Chinese authorities account information that helped convict the journalist Shi Tao, who went to prison for a decade for forwarding a government directive to a foreign website. These are documented episodes, settled in courts and congressional hearings, and they establish something the rumours often miss: the wall was not purely a Chinese invention sealed off from the world. It was built, in part, with Western engineering and Western profit. The same firms that later sold surveillance capability across borders — the trade documented in the case of FinFisher and NSO Group — were operating one end of a market whose other end ran through Beijing. Repression, as an industry, has rarely respected national boundaries; it follows the contract.

The part that isn’t a wall at all

Here is where the popular picture starts to mislead, and the correction makes the system look more formidable once you understand it. The image of a “firewall” suggests a single perimeter — a border you either cross or don’t. The more revealing scholarship, much of it by the political scientist Gary King and colleagues at Harvard, showed that the heavier lifting happens deep inside the wall rather than at its edge.

King’s team, in studies published from 2013 onward, did something painstaking: they collected millions of social-media posts from Chinese platforms before the censors reached them, then watched which ones were deleted, and inferred the rules from the pattern. What they found upended the intuitive theory. The system was strikingly tolerant of criticism of the government and its leaders — you could complain about corruption, about officials, about policy, and often the post survived. What got deleted, fast and reliably, was anything with the potential to spark collective action: calls to gather, to protest, to organise, even in support of the government. The machine, in other words, was not primarily built to protect the leadership’s feelings. It was built to prevent people from coordinating in the street. That is a subtler and more sophisticated instrument than the “no criticism allowed” caricature, and it required an army of human censors — hundreds of thousands of them, employed by the platforms themselves under legal liability for what their users posted — working alongside the automated filter.

The measurement work behind this is worth dwelling on, because it shows how much of the system can be reverse-engineered from outside. Researchers could not read the censors’ rulebook, so they inferred it — harvesting posts the instant they appeared, revisiting them hours later to see which had vanished, and building a statistical portrait of the invisible hand. A separate strand of research, by groups such as the University of Toronto’s Citizen Lab, decompiled the Chinese chat and search apps themselves and pulled out the actual blocklists of forbidden terms buried in the code, watching those lists swell and shift around sensitive anniversaries. The censorship was opaque by intent and legible in practice, because a system that must operate at the scale of a billion users leaves fingerprints everywhere it touches. That is the recurring irony of these machines: the bigger and more total they try to be, the more surface they present to anyone determined to map them.

This is the crucial mechanism the word “firewall” hides. Control of the Chinese internet is only partly a border technology. Much more of it is domestic: laws that make companies responsible for policing their own users, a licensing regime that lets the state revoke a platform’s right to exist, and the resulting culture of pre-emptive self-censorship in which the most effective censor is a nervous employee deleting a post before anyone in the government has to ask. The wall at the edge gets the nickname and the metaphor. The real weight sits inside, distributed across every company and, eventually, inside the head of every user who has learned what not to type.

The fork: from documented filter to omniscient eye

Everything above is sourced — from Chinese statutes, from court records, from measurement studies anyone can read. The mythology begins where a powerful, imperfect, well-understood system of control gets promoted into an omniscient one: the wall that sees everything, blocks everything, knows every citizen’s every keystroke, and cannot be beaten.

The reality is leakier, and the leaks are the point. Tens of millions of people in China have, at various times, used VPNs to climb over the wall — students, businesses, academics, the tech-literate young. For years the state tolerated a great deal of this, because a hard seal would have strangled the international commerce China depended on; a foreign company’s Shanghai office needs to reach its own servers abroad. The wall has always been porous by design as much as by failure, tightened and loosened as the political weather changes. It is not the flawless dome of the folklore. It is a filter with holes the state chooses when to plug.

The omniscience myth misleads in the same way the panopticon version of the Snowden disclosures misled — by confusing vast capability with total, real-time, individual control. The Chinese system is enormous and genuinely oppressive; it also relies on intimidation, self-censorship and the fear of consequences precisely because pure technical interception cannot catch everything. A regime that could truly see and block all forbidden thought would not need to jail journalists as examples or lean on companies to police their users. The examples and the leaning are the tell. The most powerful part of the machine is not a router. It is the well-founded belief, in a hundred million minds, that someone might be watching — a belief that does far more censoring than any packet filter, and costs the state far less.

What the wall is really for

Strip away the mythology and a clearer question surfaces: what is all this actually protecting? The answer the King studies point to is stability above all, with orthodoxy of belief a distant secondary concern. The Chinese state’s deepest fear, legible in what it deletes and what it lets stand, is not that people will think badly of it — they are permitted to, within limits — but that they will act together. The whole apparatus is an anti-coordination engine, built by a leadership that remembers how quickly a networked crowd can become a movement, and that watched the internet help topple governments elsewhere.

Understood that way, the Great Firewall stops being a peculiar Chinese aberration and becomes one answer — the most developed one — to a question every state now faces: what happens to power when everyone can talk to everyone at once. Other governments have reached for softer versions, whether Britain’s expanding filters or the surveillance authorities exposed elsewhere, and the difference is more of degree and candour than of kind. China simply built the fullest apparatus, wrote much of it into law, and stopped pretending the network was meant to be free. The 1987 email had promised that across the Great Wall they could reach every corner of the world. The state’s long answer was to decide, corner by corner, which reaching would be allowed — and to make the deciding so thorough, and so internalised, that most people would eventually stop testing where the wall actually stood.

What the episode leaves behind is a recalibration of the fear, in place of any tidy verdict. The instinct to imagine an all-seeing digital wall is understandable; the reality is at once less magical and more insidious. It is a system of laws, incentives and nervous human hands, engineered so that the heaviest censorship happens in the mind of the person about to type, well before the border. That is harder to picture than a wall, and far harder to climb.