Who could write protocol fiction for speculative infrastructure?

15.45, Thursday 11 Aug 2022

Writing protocols for fictional big systems might be a neat way to unlock the future. But I wonder who you would need in the room to author the spec.

To be more concrete about this, here are two ideas I’ve posted about before. How could they be bootstrapped, short of being a giant benevolent corporation?

  • What about a national packet-switched drone delivery network (2021) - there are cargo-carrying drones already. What’s the system of parcels, pick-up points, charging stations, and payment such that companies can cooperate and compete to build national delivery infrastructure, just as we have interop on - say - the phone network?
  • How about twice yearly MRIs for a personal Check Engine light (2022) - medical conditions can be spotted using computing vision on relatively cheap, non-invasive MRI scans. What’s the system of accessible scanning infrastructure and an App Store of condition-spotting health software startups, with built-in privacy and affiliate fees, such that there’s an arms race to create ever-more-comprehensive screening software and improving health outcomes?

These are both ecosystems that provide infrastructure by harnessing market forces. Get it right, and the incentives align towards getting cheaper, better, and more accessible.

So… the internet works? How did that start?

In 1969, the proto-internet ARPANET had four nodes, and it used gateways and the then-new technology of packet switching to transport data between remote computers run by different people. It was a network built to be extended. And it was.

The transport protocol was the bottom layer; higher up was the application protocol: how does software speak to remote software in a standard way. Given standards, the types of applications can flourish too. Permissionless innovation we call it now – how can you do new things without hitting coordination problems.

Two-Bit History has two fantastic articles about ARPANET:

  • The Real Novelty of the ARPANET which walks through a demo of ARPANET from 1972, what made it different from other (socially sophisticated!) timesharing networks that came before, and how its protocols were developed
  • How the ARPANET Protocols Worked which is a dive into the protocol hierarchy, how they were defined and how they worked, and a comparison with today’s internet protocols.


Protocols are just agreed ways to communicate. A protocol embodies an architecture of participation. They’re the lynchpin!

To start with there’s an enabling, base protocol which allows for (a) cooperation, (b) expansion, and (c) more protocols to be layered on top. For there it’s an iterative process as the network and its use cases grow.

To bootstrap this, a “minimum viable network” is created by a single organising body. It was BBN, under contract from ARPA, that built the enabling gateway computer (the “IMP”), and also developed its software – which included the first version of the protocol stack.

But before all of that: there’s a vision of what kind of big system is to be developed. There’s a viewpoint of how the network will grow, and what it will be used for. I think this viewpoint was developed and espoused (in the form of funding preferences and memos) by inaugural IPTO director at ARPA (1962-1964) J. C. R. Licklider. Though to check that assumption first I need to work my way through M. Mitchell Waldrop’s biography The Dream Machine, now published by Stripe Press, currently glaring at me from my bookshelves.

Applying the lessons:

When I’m thinking about a nationwide drone delivery network, or MRI-enabled medical screening ecosystem, it’s a coordination problem, right?

It would be in everyone’s interest to have these kind of big systems, but it’s in no-one’s interest to go first. It wouldn’t - for example - be in Amazon’s interest to build out a delivery drone network onto which everyone can piggyback. That’s a catalyst problem.

In broad brushstrokes then, we want a process like this:

  1. Imagine the destination ecosystem such that its internal economy is well-balanced, there is good interop, and the incentives are all pointing in the right directions
  2. Draft a protocol – a set of rules for how to work together, but expressed as rules primarily for software. Ensure that it represents the desired ecosystem, but also that it works when there are a small number of actors, and that it contains incentives to grow without introducing coordination costs
  3. Launch a prototype: a working reference implementation for the protocol, but also something that achieves commercial end-to-end
  4. Smooth the way for growth by removing blockers for 3rd party participants (for example: underwriting risk on long-term infrastructure investment)
  5. Iterate everything.

The protocol is where the rubber hits the road. It’s a description of the future, and a proof of the potential economics. If done well then funding the prototype should be a relatively straightforward public infrastructure decision – although there may need to be policy whitepapers to communicate the cost/benefit to government…


In the adjacent world of software, Robin Sloan has been considering the problems of Twitter and social media - everything from the doomscrolling user experience to the failure modes of centralisation - and has come up with a vision of something new.

But he hasn’t built software. He has designed a protocol: What follows is a narrative description of a protocol that I believe might open up some interesting new possibilities on the internet.

In depth!

Spring ‘83 is a protocol for the transmission and display of something I am calling a “board”, which is an HTML fragment, limited to 2217 bytes, unable to execute JavaScript or load external resources, but otherwise unrestricted. Boards invite publishers to use all the richness of modern HTML and CSS. Plain text and blue links are also enthusiastically supported.

(It’s enormously readable. Check it out.)

There’s a draft spec! github.com/robinsloan/spring-83 – and, since the ref was first published, various people have created various implementations, also listed at that link, so you can use it too.

The thing is that, for an ecosystem, you do need many participants.

With the narrative description, Sloan created the catalyst. With the spec, he solved the coordination problem.

Let’s pretend we wanted to write the protocol for a nationwide, interoperable, drone-delivery network.

Who would you need in the room to kick off that process?

Thinking out loud, you need expertise in at least these areas:

  • Existing market dynamics around delivery
  • Existing market dynamics around drones
  • How to credibly describe a future industry sector (for example: who planned out 3G before anyone bid for the spectrum?)
  • The theory of protocol design and what allows for attributes like growth, interoperability, and permissionless innovation
  • The practicalities of protocol design: how to be simple, readable, adaptable
  • Contract engineering with subject-matter expertise - to consider how to build the reference implementation
  • Operations and fundraising - to consider how to run and scale the reference implementation
  • Storytelling: visualising the future and how we get there. Vital for alignment and enrolling support, but also (the vital work of design fiction) to test, in thought experiment form, against unintended consequences
  • Lobbying: how to write policy whitepapers and get them discussed by the right people
  • Lastly, a convening function, from facilitating to goal-setting

Anyone else?

(That’s too many people. So perhaps with the right economist, and the right technologist, etc… You’d want a small group, I think.)

Output: a report with an imagined market, some kind of visualisation, a designed and documented protocol, a costed approach for the prototype build, and a wrapper for all of the above to carry it into the right audiences (public sector; VC; etc).

Ahead of that, you would want probably want to kick off discussions with a pool of people who are both open-minded and also well-networked in the above areas, in order to iterate to the right group members.

And beyond the delivery of the above protocol fiction etc by my new Committee for Actualising Speculative Infrastructure, an organisation to carry the ball forward.

Back in 2020, I spoke at ThingsCon about wanting to work at imagining beyond design fiction:

But we don’t need just design fictions. We need business model fictions, engineering feasibility study fictions, interop protocol specification fictions, investment return fictions.

This is what I meant!

Although I seem to have drifted from protocol fiction to committee fiction…

But 50% haha-lol-what-if and 50% seriously: if we got half a dozen people together in London sometime, who should be in the room?

If you enjoyed this post, please consider sharing it by email or on social media. Here’s the link. Thanks, —Matt.