Where Wizards Stay Up Late

The book exists partly to displace two persistent myths: that the network was built to survive a nuclear attack, and that any single person invented it.

Their account is that it was built to share scarce and expensive computing resources between research institutions, and that survivability was a property of the design rather than its motivating requirement.

It was not built to survive a nuclear war. It was built because computers were expensive and there were not enough.

Devmark’s reading — not a quotation from the book

The foundational technical idea: break messages into packets, route each independently, reassemble at the destination — as opposed to establishing a dedicated circuit for the duration.

Their account of its independent invention by Paul Baran in the United States and Donald Davies in Britain is a good illustration of an idea arriving when the conditions for it exist, and of how differently the same idea was received in different institutional contexts.

The engineering heart of the book: the decision to put a separate small computer at each site to handle networking, so that host machines did not all need to speak the same protocol.

This is the architectural decision that made the network buildable, and the account of Bolt Beranek and Newman delivering the first ones on an impossible schedule is the book's most engaging narrative section.

Their account of how the protocol design process worked: a young graduate student writing up a proposal, hedging its authority by calling it a request for comments, and inadvertently establishing a norm that persisted.

The RFC culture — open, informal, deliberately non-hierarchical, judged on technical merit — shaped the internet's governance for decades, and its accidental origin is one of the better stories in the history of standards.

The most durable governance norm in computing started as a graduate student hedging about whether he had authority.

Devmark’s reading — not a quotation from the book

Their account of electronic mail emerging as an unplanned application, quickly becoming the dominant traffic on a network built for resource sharing.

The lesson they draw — that the most consequential use of infrastructure is frequently not the one it was designed for, and that general-purpose designs permit this while optimised ones do not — is the book's most transferable observation.

Their portrait of the working culture is the book's other subject: collaborative, informal, remarkably non-proprietary, and organised around rough consensus and working code.

Their argument is that the network's open architecture followed from this culture rather than from a technical requirement, and that a differently organised group would have built a differently shaped network.

A theme handled without much editorialising: the network was built with military research funding, by civilian researchers, with a great deal of latitude.

The book neither celebrates nor condemns this, and the resulting picture — that a deliberately hands-off funding model produced something nobody specified — is more interesting than either position.

It ends before the web, before commercialisation and before the internet became a consumer technology, which means it covers the origin and not the consequence.

The account is also predominantly American, treating parallel European work as background, and its interview-based method means it reflects the recollections of the participants who were available and willing.

Well. It is a genuine piece of technology journalism with a real talent for making engineering decisions legible to a reader with no background, and the personalities are drawn sharply enough to carry that reader through material that is technical in places.

It is also comparatively short, which is unusual for this genre and welcome — the authors resist the temptation to pad the origin story out to the length of a general history of the internet.

The general-purpose argument: a system designed for a specific use supports that use, while one designed to be general supports uses nobody anticipated, and email is the demonstration.

And the RFC norm, which is worth knowing about as an example of how a governance culture can be established almost by accident and then prove extraordinarily durable.