The way to Platformland

Platformland: An anatomy of next-generation public services
by Richard Pope. 282pp, London Publishing Partnership, September 2024

Richard Pope’s Platformland is a vision of the state as a set of interlocking, cooperating software systems, a refinement and extension of Tim O’Reilly’s idea of Government as a Platform. In Government as a Platform, O’Reilly built on Eric Raymond’s The Cathedral and the Bazaar, proposing that the state enable a rich ecosystem of civic software like the one Raymond described for open source. In Platformland Richard Pope builds in turn on O’Reilly’s work, setting out the pieces of the state’s ideal “stack”—credentials, rules, services, components and data—and fitting them together in 10 chapters, one idea per chapter.

The technology Pope’s vision depends on is all old and boring, in a good way: relational databases, message queues, logs and so on. What’s new is that digital government practice in the UK is now mature enough to think about a world where we focus on, in Pope’s words, “the systematic elimination of administrative burden”. Platformland’s alternative to incremental “digital transformation” is to deliberately fund the creation of infrastructure to support that systematic elimination. If we don’t do this, that infrastructure won’t emerge organically from government agencies because, as Pope puts it, the institutional incentives are simply not strong enough.

Pope finds inspiration for this vision in well-functioning digital systems of government around the world. He asks quite reasonably why Britain, where GOV.UK once led the world in civic technology, can’t join up bureacracy as well as other countries do.

Raw materials

Out of the 10 ideas, “credentials” are Platformland’s most brilliant. Pope proposes that a credential is proof of something, that the output of a process can be a credential, that a credential can be used as the input to other processes, and that a credential is not the same as an identity. This fine distinction between uncountable “data”, fragile “identity” and purposeful “credential” beautifully opens up the potential for interoperability, because it’s a much clearer way to think about the information services need. Pope makes a significant leap when he points out that digitised credentials are “the raw materials to automate eligibility”. From this beachhead he’s able to construct scenarios of “proactive” services that delete whole categories of bureaucratic activity. Later in the book he goes on to conscientiously consider the human costs and merits of machine-mediated interactions (tl;dr: it depends).

The act of naming, of “anatomy” (Platformland’s subtitle is “An anatomy of next-generation public services”) is the foundational act of software design, and it’s characteristic of Platformland that “credential” is an extremely well chosen technical name. It abstracts over “official document”, “qualification” and “proof of ownership”. It also carries useful connotations of authenticity and of verifiability. And of course “credential” goes on to enter the vocabulary of the book, where it becomes a building block in further proposals.

Pope always credits his intellectual precursors. In the case of credentials he builds on the concept of “identity unbundling” from Hal Abelson and Laurence Lessig, and deftly introduces Jyri Engeström’s concept, new to me, of “social objects”: “the manifestation of the relationship between a service and its users”. Frankly, it’s exciting when big ideas like that couple with concrete proposals to do useful things like “automate eligibility”. And whenever he calls on practices or principles, whether that’s user centred design, relational databases or the expression “it just works”, Pope takes the trouble to mention the origins and to set those particular ideas in context. You get the sense that he is allergic to dogma and obsessed with getting to the bottom of things. And the book is fun because you are never more than a few pages away from an illuminating bit of lore. Platformland’s combination of rigour and lightness of touch feels clever and unusual.

Seams

Whilst his anatomy of ideas is always meticulous, Pope is often happy to leave other things vague. For instance, the relationship between “credentials” and another part of the stack, “data”, is fuzzy: at one point we imagine an “audit office” which verifies that updates to a register of land ownership (“data”) are valid. But isn’t a claim of land ownership a credential?

It doesn’t really matter, because this is only a taste of the complexity that will confront would-be implementers. But implementation can’t be ignored. There is a chapter on “seams”, meaning services designed with openings we can reach into to find accountability. (In a pleasing rebuke to people who would cargo-cult “Silicon Valley minimalism”, Pope observes that those “seams”, necessary for good quality public services, are exactly what the iPhone doesn’t have). For me, “seam” evokes the work of Michael Feathers, where the seam is a boundary to find between parts of a codebase, a way to hold one part of a system steady while you straighten out another. Whilst the language of software architecture is everywhere in Platformland, the language of refactoring isn’t, although refactoring is how architecture comes into being. Perhaps that’s because refactoring is by nature a process of discovery. Bringing these ideas to life will require some highly context-dependent choices about what to salvage and what to rebuild.

Implementation questions are out of scope for Platformland. I appreciate that to include them would weaken it. But that doesn’t mean they aren’t important. The larger scale the architectural change, the riskier and the harder to deliver it becomes.

Rats’ nests

Sometimes you find yourself agreeing so vigorously with Platformland that you want to fling the book across the room shouting I KNOW! YOU’RE RIGHT! IT’S JUST HARD, OK? Because basically, Richard Pope is right. Platformland would be a better place to live. The depth and completeness of his vision is clarifying and inspiring. But for those who would make it happen, the question of how we get there is simply not separate from the question of where we want to go.

One of Platformland’s core provocations comes in the chapter about common components:

Conducting digital transformation service by service, form by form, one casework system at the time is not the answer. It might be great for consultants, but it will mean that the heat death of the universe will arrive sooner than every service being brought up to 2005 standards.

Yes. And yet. The databases are not well structured. The outsourcing contracts are locked in for years. The domain knowledge to unpick the information is scarce. Building platform tooling is relatively easy, but Platformland’s furnaces run on information locked up in Excels and proprietary CRMs, deep in rats’ nests of omnidirectional coupling. We can’t just stop putting one foot in front of the other to bring those systems into line… can we?

Platformland challenges us to reform these systems more swiftly and more radically. Revolution, not evolution again, and the original revolution still unfinished. That’s a provocation.

How?

Pope ends with two questions that feel rather academic after the hard design proposals that make up most of the book: what should the bounds of the public sector be in the digital age, and what will ’the politics’ of digital services and infrastructure be?

I think the means to answer both of those can only be found deep in implementation. So I want to pose some more concrete questions.

Firstly, from amongst the blunt instruments of policy, re-org and procurement, what levers can we use to enable meaningful experimentation and implementation on the scale Platformland requires? How does the Platformland agenda get the political momentum it needs to enable its reforms? And on what time horizon shall we expect to make progress—years, a decade, two decades?

An example of the complexity Platformland faces: the book deals with issues bordering on a particularly British problem, digital identity. This is a political hot potato but also a critical underpinning of the data ecosystem Pope lays out. You could pull a set of principles out of Platformland for a system of “identity” which sees off “civil liberties” objections while also supporting the necessary information flows. But getting there technically and politically will be a long and uncertain road.

If not now, when?

(If you’ve read this far, you should buy the book. There’s a huge amount in it that is good that I haven’t mentioned here.)

In Government as a Platform, Tim O’Reilly quoted John Gall’s Systemantics:

A complex system that works is invariably found to have evolved from a simple system that worked. The inverse proposition also appears to be true. A complex system designed from scratch never works and cannot be made to work. You have to start over beginning with a working simple system.

I love this quote and I think it’s true. It’s the key challenge to Platformland’s call for comprehensive reform over incremental change. Many non-technical readers of Platformland, convinced by its vision, will be more than happy to nod along with that provocation. Maybe they’ll even make some rash decisions off the back of it. Technical readers are more likely to be thinking along the lines of John Gall, wondering how all this could possibly be made to work.

I wish I had a Richard Pope-style intellectual precursor for the feeling that we all tend to radically underestimate the natural complexity and resilience of systems we want to reform. With good reason, we see starting from scratch as extremely perilous. But I believe that makes us discount the idea too readily. So whilst top-down reform on a national scale is obviously the road to hell, I think we should take Platformland’s invitation to radicalism seriously.

What would a bold, progressive refactoring of the digital state involve? A reasonable starting point might be a pilot whose scope cuts across multiple government departments, supported by a mandate for fusing organisational reform and technical refactoring. (Pope’s ideas of “abstract” and “composite” services that combine other services are appealing). Perhaps we could see what it would take to apply the once-only principle, which prohibits government from collecting the same data twice, in a particular domain. That means getting out of the centre and deep into departmental business, reaching into those broken, heavily coupled systems, and running the first ropes between them—probably without much regard for backwards compatibility.

That process would upset traditional data analysis functions, interfere with contracts, impede progress on existing initiatives, disempower teams who are trying to deliver incremental changes, and irritate departmental leaders, if it gained traction at all… and that’s just to get started.

But if not now, when?