I've been thinking about centralization for a long time.
Years ago, I thought I understood the direction fairly clearly.
Large centralized systems were becoming brittle. Networks were making it possible to distribute capabilities that once had to live in one place. Eventually, I thought, the same organizational change we had seen in computing would appear elsewhere.
Some of that happened.
Some didn't.
The analogy turned out to be more useful as a question than as a prediction.
Computing provides an appealing story about decentralization.
Early systems could place enormous capability in a central machine while terminals around it depended on that machine to do most of the work. Later, networks made it practical to distribute computing across many machines.
That brought real advantages.
Work could happen in more places. Capacity could be added differently. Some failures could be isolated rather than taking down everything around them.
Then we built enormous data centers.
Apparently the lesson wasn't simply that centralization was obsolete.
The internet itself contains the contradiction.
It is distributed enough that information can move across many networks, yet much of what we actually do online depends on a relatively small number of companies, platforms, data centers, payment systems and other intermediaries.
Both structures exist at once.
Once I noticed that, the question became more interesting.
Maybe centralization and decentralization aren't opposing destinations.
Maybe they're tools.
A town probably shouldn't manufacture its own microprocessors.
It might want more than one way to communicate during an emergency.
A small business doesn't need to operate its own payment network.
It may want to retain enough of its customer relationships that losing access to one platform doesn't make the customers disappear.
A community can benefit enormously from a large electrical grid while still finding value in local generation that can continue operating under some conditions when the larger grid has trouble.
The question changes depending on what we're trying to protect.
Centralization is very good at some things.
It can establish common standards, pool expensive resources, coordinate activity across large distances and make complicated services available to people who couldn't reasonably reproduce them locally.
That's why we keep building centers.
They solve real problems.
They create some too.
As more activity passes through a central institution, failure there can affect more people. Decisions become farther removed from some of the places experiencing their consequences. An organization can become efficient at the scale it measures while creating costs somewhere outside the measurement.
Distribution changes those tradeoffs.
Local systems can preserve capabilities when larger ones fail. Different approaches can coexist. Someone closer to the consequences may be able to notice problems sooner.
Then another set of difficulties appears.
Who maintains the standards?
What happens when two systems need to communicate?
Who pays for duplicated capacity that spends most of its time unused?
What happens when a small local institution is badly governed?
Distance isn't the only place accountability can disappear.
Anyone who has lived in a small town knows that proximity doesn't automatically produce wisdom.
So I've become less interested in decentralization as a principle.
I'm more interested in deciding where a center is useful and where dependence on that center becomes dangerous.
There are some practical questions that help.
What happens if this institution disappears tomorrow?
Can another part continue functioning?
How difficult is it to leave?
Who holds the information needed to rebuild the relationship somewhere else?
Does concentrating this function create a capability we couldn't otherwise have?
Does distributing it preserve something we'd otherwise lose?
And who experiences the consequ