Studying network processors and virtual machines with Jennifer Rexford at Princeton and software-defined networking at Stanford in the summer with Casado, Ali became a serious scholar of the practice and philosophy of moving computer and routing functions from fixed hardware to programmable software. Casado went on to found the pioneering network virtualization firm Nicira, which he sold to VMWare for $1.2 billion. Ben Horowitz, the son of the truth-telling provocateur David Horowitz, had made his fortune inventing software at VMWare, and Casado eventually joined him as a venture partner at Andreessen Horowitz.
Whether SDN or network function virtualization, the virtualization movement that engulfed Ali was in the process of transforming the network. It changed from a seven-layer structure dominated by hardware capabilities to a two-layer structure largely defined by software mimicry of hardware capabilities. Like Stephenson’s Hiro and the pseudonymous Satoshi, Ali lived in a realm where it was possible to break away from the limitations of the material world into the electronic night and create a metaverse that fulfilled your dreams.
The seven-layer model consists of a hierarchical stack in which lower functions are controlled by higher functions. At the bottom is the physical layer, the fiber-optic lines, microwave oscillators, mixers, 1550- and 900-nanometer lasers, photodetectors, silicon routers, erbium-doped amplifiers, and twisted-pair telephone wires, antennas, coaxial cables—the list is endless—that carry the data packets across the network at the behest of the layers above it. Hard to design and build, this layer of hardware devices is at the heart of the miracle of modern electronics. But by the time Ali was studying at Princeton, much of the industry was ignoring hardware to build Turing machines in the ether.
To understand the contemporary Internet, you have to take these hardware miracles for granted and build castles in the sky—in computer language, “stacks”—that can imitate hardware and transcend it in virtual threads and cores and chains. But the evolution from micro-matter to metaverse begins with the seven-layer netplex scheme of the Open Systems Interconnection (OSI) model of the International Standards Organization.
In the OSI stack, above the physical layer is the datalink. This is the medium where hardware becomes “firmware” and software that define the electrical specifications, timing rules, and electron-photon conversions that enable the transmission of information across a link from one node or computational address to the next. Switches operate here at layer two, passing packets only to the next node. Local area networks such as Ethernet or WiFi function at this level. If you shun the highways of the Internet, you can live your life transmitting your bits and bytes across the datalink layers, layer two.
The third layer is the network layer, the domain of routers, which combines with the transport layer (layer four) to establish the end-to-end links that constitute the TCP/IP Internet Protocols. This is the entire system of IP addresses and Transport Control Protocol traffic shuffles that comprise the connections from end to end across the Net. Layer three does the headers on the packets, the identities and addresses; layer four does the actual transmission and reception of data packets and traffic management, load balancing and acks (I got it!) and naks (I’m still waiting) that assure connections. Layers three and four tend to be a bastion of central powers, where governments and their intelligence arms chase down domain names and addresses, entities such as ICANN and even the UN’s ITU. When they discover a Silk Road or perhaps an Alpha Bay, they track it down over layer three.
Above layer four is layer five—the all-important session layer—which governs a particular two-way communication from beginning to end, whether a video stream, a Skype call, a Session Initiation Protocol conference, a messaging exchange, an email post, or even—and this would prove fateful—a transaction.
Layers six and seven are the schemes for presentations and applications—user interfaces, windows, formats, operating systems, and so on. These are summed up in the ingenious schemes of hyperlinks (click on a word and go to a new page) and universal resource locators (URLs) addresses. Tim Berners-Lee at CERN in Geneva invented them in 1989 as part of his World Wide Web. Berners-Lee wanted to make all data linkable into one Web, a skein of tools that made it easy to set up a Web page of “shared creative collaborative space where everyone could play together.”
As 70 percent of all links came to be handled through Google and Facebook, Berners-Lee feared that his Web was dying. He would become a Blockstack enthusiast. “When he heard what we were doing, he did a little dance,” said Blockstack’s software chief, Jude Nelson.
To describe the OSI stack in Telecosm, I used the example of a phone call. Pick up the handset and listen for a dial tone (physical layer signal), now often simulated; dial a number (every digit moves the call another link closer to the destination); listen for the ringtone (signifying a network connection and transport of signals). By the time you get someone on the line, you have gone through the first four layers of the OSI stack. Then your “hello” begins a session, the choice of English defines the presentation, the conversation constitutes the application layer, and the hang-up ends the session.
While a materialist might suppose that the physical layer is all, and a software triumphalist imagines that it is all in his head, the genius of networking is dualistic. Animated by trillions upon trillions of microchip transistors, vias, and traces, the physical layer is ultimately as opaque and unfathomable as it is ingenious and indispensable. Software logic proliferates in a hierarchy above and defines what the hardware does.
As every component speeds up in accord with Moore’s Law, many of those special-purpose devices—ASICs, networking chips, network processors, TCP accelerators, traffic managers, and routing lookup-table content addressable memories—are less needed. Replacing them is increasingly fast, dense, and programmable general-purpose hardware.
Substituting for custom devices in routers, switches, and other networking gear are powerful servers based on multicore general-purpose microprocessors from companies such as Intel, Cavium, and Mellanox. They link together under the guidance of ever more complex and integrated software. General-purpose hardware commanding vast markets across the industry—from billions upon billions of smart phones and video game consoles—has become ever faster and cheaper. In time these chips could displace the arrays of more costly specialized hardware previously necessary to perform trillions of operations per second at fiber-speed across the Internet.
With the right software, an Intel Xeon microprocessor in a fast server could perform router and switching functions that previously required elaborate custom hardware from Cisco with names like Tiger and Quantum Flow or ingenious fiber-speed network processors from Israel’s EZchip/Mellanox.
In the end, Google shunned most specialized networking hardware in favor of thousands of servers deployed across vast datacenters and integrated by software. The Turing machines were as immaterial and mutable as they were in Turing’s mind. A router or computer or switch or Internet realization could be “virtualized” and have no specific hardware manifestation at all.
Leading this change were people like Casado, Rexford, Freedman, Horowitz, and hundreds of others across the industry. These network scientists introduced Ali and the other Blockstack inventors to blockchain engineering of these principles. They scrupulously separate the control plane on a higher level from the data plane on a lower level. This design assures that these architectures are uniquely streamlined and scalable.
It all began with that first computer in Pakistan, a beguiling prize once he had mixed and matched the components and assembled it from the kit. He recalls that when he had completed the job, though, the computer left him puzzled. In Pakistan at the turn of the twenty-first century, a computer was like the proverbial “car in the jungle.” A car might offer attractive features—light, heat, air-conditioning, shelter, protection—but a car becomes really exciting only in conjunction with roads. Ali’s computer fully enthralled him and changed his life only when he acquired a Netscape
browser and went online. There in Pakistan he could range the roads of the entire World Wide Web and become a citizen of a global information economy.
As Ali sensed, the rise of Netscape marked a turning point in the history of networking—the provision of new accessible roads for data. Its browser provided interactivity, text, imagery, security, and transaction possibilities across the Web. It embedded Brendan Eich’s JavaScript for dynamic Web pages and transactions forms, a secure-sockets layer enabling safe commercial links across the Net, and a Java Virtual Machine to port apps from any of the Tower-of-Babel of operating systems.
Netscape’s founders saw the Web as an arena for interlinked creative expression of all kinds, from photos to videos. Its founder, Marc Andreessen, and the investor Jim Clark, inventor of the 3D “geometry engine” at Silicon Graphics, both anticipated a 3D metaverse of games and virtual worlds. Through Netscape, Andreessen, Eich, Clark, and colleagues had given Ali the power to animate Web pages, share them with the world, and possibly make money on the Net.
The Netscape IPO in 1995 also meant distribution of the rewards of the Internet. On the very first day, the shares almost tripled to a valuation of more than $3 billion, benefitting the public while inspiring and funding entrepreneurs to challenge the computing establishment. In the following five years, a spate of IPOs, from companies such as Google, Amazon, and a thousand dot-coms, fueled a boom of distributed Internet applications. Under what I called the Law of the Microcosm, innovation moved decisively to the edges of the network.
It was a high point for technological entrepreneurship. After the year 2000, however, the number of startups would stagnate and IPOs nearly disappear for all but the biggest tech companies. In the wake of the Enron debacle, regulations under the Sarbanes-Oxley Act exacted a toll of some two million dollars for reaching public markets and imposed a rigid accounting regime high on paperwork and low on trust. It was entirely inimical to startup culture and finance.
Typical of the folderol that makes it prohibitively costly and treacherous to be a public company is the “fair disclosure” lawyerization of all company communications. If you have to pass it by the lawyers, you probably won’t say anything of interest. All but the largest corporations became realms of nearly zero-entropy communications—all retro-numbers and no inside detail to make them significant.
By the time Ali arrived at Princeton in 2012, Netscape had foundered. Its browser gave way to Microsoft’s Explorer, rolled out free and bundled with Windows 95. Initiating the now common practice of Internet leviathans’ buying innovation, Microsoft was able to quell the Netscape challenge by acquiring the Spyglass browser. The leading designers of Spyglass, as it happened, were Netscape’s Andreessen and Eric Bina, who had developed its basics as Mosaic while at the University of Illinois’s supercomputer center. Microsoft acquired an elegant modular browser and made Netscape’s inventors compete with themselves.
The IPO dearth continued for more than a decade. For nine months in 2016, there were no US IPOs at all. Instead, venture capitalists kept hundreds of “unicorns”—private companies valued at more than a billion dollars—in their corrals. Led by Uber and Airbnb, nearly all of them had private market caps higher than Netscape’s in its IPO. Most were less interested in going public than merging with a mammoth like Google/Alphabet or Facebook. Unlike the appreciation of earlier Internet companies such as Microsoft and Netscape, the appreciation of unicorns would not chiefly benefit the public. The returns (and burn rates) flowed mostly to the venture capitalists who held them and the leviathans that bought some of the best of them.
This was the situation when, in 2012, Ali and his friend Ryan Shea joined the Entrepreneurship Club at Princeton and together plunged into an effort to launch new Internet apps. By the spring of 2013, they found themselves strangely stymied. The Internet roads they were traveling now converged in giant data center hubs that offered scant security or privacy and little economic yield for any but a few Internet giants.
It was an enclosure movement with a critical flaw. An insecure Net could not protect property rights, defend privacy, host safe and efficient transactions, permit micropayments to halt spam, or establish sure identities. Google, Facebook, Amazon, Apple, and the others responded with proprietary “safe spaces” of their own. There they could accommodate commerce among their mostly locked-in users.
As Ali writes, “Currently, with frequent use of an online service, user data gets locked into ‘data-silos,’ e.g., data that is understood and stored by Facebook, Yahoo!, Google and others respectively but cannot be migrated across services. This leads to a centralized data model; the data silos inevitably get hacked eventually, e.g., the recent hack of 500 million Yahoo! users.”7
These silos, or “walled gardens,” were what depressed Berners-Lee.8 They worked well for their proprietors but destroyed the global coherence of the Net and caused increasing segmentation. Within the segments, Google, Apple, Facebook, Amazon, et al. collected more and more private data and protected them with firewalls and encryption. But as time passed, they discovered that centralization is not safe. Putting data in central repositories solved hackers’ hardest problem for them: It told them which data were important and where they were, putting the entire Internet at risk.
Google mobilized “an all-star hacker swat team” to strike back at dark-side hackers. An entire industry of security firms emerged to protect the user data honeypots by reacting to outbreaks of viruses, grand data thefts, denial-of-service attacks, malware, malvertisments, phishing schemes, ransomware, and other mischief. Each Internet fiefdom responded by foisting on its customers a flurry of security busywork that did nothing to improve security and got worse every year by every measure. “Security” programs merely let butterfingered data holders tell the courts that they were doing all they could, pointing to their enormous expenditures on such programs.
The leviathans’ data silos inspired tyrants around the world to isolate their own Internets. If two nerds at Google could have their own Internet and one boffin at Facebook could have his, why not one for the Chinese government? Or the Iranian mullahs? Or, heaven forfend, the European Union? Google would hear from all of them.
The Internet stack had become a porous and perforated scheme in which most of the money and power could be sucked up by the big apps at the top run by companies such as Google. What was needed was a blockstack that could keep the crucial IDs and personal data and pointers to storage addresses in a secure and immutable database on the blockchain.
As Ali and Shea understood, security is not an app or a video game. It is an architecture. Resolving to design that architecture, Ali became an American citizen and—with Brendan Eich, Vitalik Buterin, and other pioneers—a leader of the movement to reestablish the Internet on the decentralized, peer-to-peer principles that he experienced as a boy in Pakistan.
CHAPTER 15
Taking Back the Net
Muneeb Ali, Ryan Shea, and their team were ready to defy the Google model and restore a centrifugal Internet. They were proposing the breakdown of the system into just two key structures: monolith, the predictable carriers of the blockchain underneath, and metaverse, the inventive and surprising operations of its users above. They would provide the foundation for a creative cornucopia.
Continuing to work on his “New Internet” Ph.D. thesis but running out of funds again, Muneeb Ali was more interested in actually building the new network. In 2012, he had the good fortune to meet Jude Nelson, a tall, blond-haired software engineer from Arizona who needed help in Manhattan on a storage project named Syndicate, an enterprise file-storage system based on blockchains. Using cryptographic mathematics—hashes across time—it provides safe pointers to computer storage locations and addresses. Syndicate could use storage facilities like Google Drive, Amazon’s S3, and Microsoft Azure as utilities by storing pointers and ID in the blockchain, with the data’s owners retaining control.
Nelson and Ali felt a surge of excitement about blockchain technolog
y. Ali described it as “the most sophisticated and complex and yet elegant and beautiful program I ever came across. And the main thing it does is it gives power back to the people.” He joined Nelson working on Syndicate, and two years later Nelson came to work for him in Manhattan at Blockstack, then called “OneName,” which Ali and Shea, on leave from Princeton, had started in 2013.
“Apps were not responsive to customers so much as designed to lock them in,” says Shea. “You go onto the net and Facebook or Google or Dropbox or Pinterest or Amazon and all want you to move in, giving them all your documents, music, providing storage for your life. Medical sites want to store all your health data. You have to petition to get it when you want it”—when you need to move to a new provider, for example.
The Blockstack team wanted to reestablish the network on reliable, low-entropy foundations. Ali explains: “Decentralized identity systems enable users to control a unique identity recorded on the blockchain that can be recognized by any site.”1 He contrasts this universal ID with the current “username and password combo that can only be recognized by the site that had you create an account.” With the blockchain, users can log in to the websites by automatically proving ownership of their identity.
When Ali and Shea were getting started on their project, thousands of new Internet companies were forming around bitcoin and the blockchain. But most of them were supplying new services (Abra’s blockchain-based remittances to Third World countries), currencies (Monero and Zcash, with crypto-coins for ever-deeper privacy), forums (Steem, a blockchain Reddit news and commentary board), and markets (AlphaBay, emulating the Silk Road drug bazaar).
Life After Google Page 17