Tim Berners-Lee and the web I still want

I met Tim Berners-Lee at Davos last month. We spoke for about fifteen minutes in a corridor between sessions, which is where many of the useful conversations at this gathering happen. He had just come from a panel. I had come from a meeting about newsroom technology. We ended up talking about what has gone wrong with the web he invented, and what a second chance might look like. My thanks to Chris Merritt at Cloudflare.

I opened my first website at rajiv.org in 1995. By the time the browser wars of the late 1990s gave way to the platform era, I was running newsroom engineering teams at American newspaper companies. The web had given me a place to publish and a career building places for others to publish. Meeting its inventor, I wanted to talk about how to keep those possibilities open.

The drift from pages to platforms

The early web let authors host pages where they chose and link to one another. I could put something on my own site and give someone its address. A reader did not need to join the same service before following the link. That freedom was part of what drew me to the medium.

Much of our publishing and conversation now takes place inside services that hold our data and control how we reach one another. Leaving can mean losing access to an audience, an archive or an identity built over years. Even when a service lets you download your records, it may offer no way to keep the relationships and activities that made them useful. That dependence is what Tim wants to undo.

I had heard him make this argument in public talks. In person, he talked about household data: calendars, contacts, health records, where a person has been and what they have read. He wants that information to live in a space the user controls, with applications asking permission to use it.

When I asked why he thought the timing was better now, he pointed to changing expectations about data rights, including GDPR, and to exhaustion with breaches, intrusive design and surprise policy changes. He thought pressure from users could make developers reconsider a business model built around holding their data. Health was one area where he expected that pressure to be strong. It is an adoption argument I take seriously: the people who benefit from control have to want it enough to choose it.

An address book you can keep

In his September essay, One Small Step for the Web, Tim describes separating the choice of where to store information from the choice of applications that use it. Solid is the open-source project developing that approach. A person keeps records in a web-accessible data store, called a pod, and applications request access. Inrupt, the company he co-founded, is working to support the project commercially. This leaves room for businesses that provide storage and businesses that build applications, without requiring them to be the same business.

Suppose you find an address-book app you like better. You want a different way to search and edit your contacts. You aren't trying to rebuild your contact list. Essam Mansour and his coauthors' 2016 demonstration paper describes two contact applications working on the same set of records. Their plan treats this separately from moving a user's storage between providers.

You could use one compatible app to find a contact, then another to edit the same record. Once the edit is saved, the first app can retrieve the changed record from the same place. There need not be an export file between the two apps, or two independently maintained address books to reconcile. Trying a different interface would no longer have to mean leaving the place where your records live.

A compatible app needs to find the records and understand their structure. Solid's specifications describe how information is represented and how applications read and write it using web requests. Those conventions give developers something to build against. They do not make an arbitrary address-book app understand another app's data.

Imagine that one app lets you attach a note to a contact and another has no place to display it. The second app should preserve the note when you edit the email address. Otherwise, a choice that looks harmless on screen could discard information. Being able to read the same document is a starting point; preserving its meaning through an edit is the more demanding test.

Solid's access-control specification distinguishes reading a resource from changing it, and both from changing its permissions. An app used to look up an email address has a different reason for access from one used to maintain the address book. I would rather make that decision in the context of the task than approve a broad request whose consequences are unclear.

Once an application has received a contact's email address, changing permissions at the source cannot erase a copy it kept elsewhere. A permission change can stop subsequent requests that are no longer authorized. The app's conduct after receiving information still matters. Choosing where the data lives does not settle whom to trust with it.

If you are happy with your pod provider, trying another compatible app need not disturb that relationship. If you want to leave the provider, the data still has to go somewhere. The 2016 paper's migration example includes moving the pod and changing the user's profile to point to the new storage. I would want to know how to recover the records, move them, and keep the applications finding the right place after a move.

I would judge a Solid application by following a contact through the whole process: open it in a second compatible app, change a field, and check what both apps preserve. Then look separately at the storage provider's exit procedure. A download button answers whether I can obtain a copy. I also want to know what I can keep doing with it.

For a developer, this is an attractive constraint. The interface has to earn its place even when a user can try another one against the same data. A better way to find a contact or edit a crowded record can be enough reason to choose an app; the developer need not also persuade the user to relocate the address book.

What this could mean for journalism

Most of my questions to Tim were about news. Every newsroom I have worked with in the last decade has felt the squeeze of platforms sitting between publishers and readers. They set distribution rules and change them. They capture advertising while the newsroom remains responsible for the journalism. A reader's desire to follow a writer or a beat becomes entangled with the platform's decision about what to show next.

I asked what Solid could offer news organizations. The possibility he described was a reader's interests, subscriptions and reading history living in the reader's pod. A reader could carry that information between services and decide which publishers could use it. Publishers could personalize from what the reader chose to share, rather than having to assemble a profile through surveillance. Recording a subscription in a pod would not by itself make a paid entitlement valid at another publisher; that would require the publishers to agree. The reader could carry that identity to each participating service.

I am not sure this will happen in the form he described. News organizations have spent years trying to reclaim a direct relationship with readers while platforms have gained scale. A different data model alone will not resolve the commercial incentives. Still, the direction matters to me. Publishers could compete through the journalism and experience they provide, and readers could have more say in the terms of that relationship.

The web I still want

Tim is asking for the agency of the early web on the infrastructure of 2019. His approach gives developers specifications and implementations to build on, alongside Inrupt’s commercial support and his plans for its governance. The engineering makes the ambition something other people can contribute to.

The original web spread in part because others could build on it freely. CERN's decision to put the software in the public domain in 1993 helped make that possible. I want the next generation of web applications to preserve that freedom for the people using them as well as for the people building them.

What stayed with me was Tim's persistence. He is in his sixties and could spend his time being celebrated for what he already made. Instead, he is still working on what he thinks the web should become, quietly and without the showmanship that surrounds many figures of his stature. I came home wanting to keep building toward that web: one where readers and publishers can choose how they meet, and where changing an application does not mean surrendering a part of your life to it.

Technical references: Solid Specification, January 18, 2019; Solid Web Access Control specification, version 0.5.0, November 29, 2018.