It is hard to change constitutions – deliberately so.  It is hard to re-engineer physical infrastructure – intrinsically so.  It is hard to stop and start again from scratch.

Every decision and every context in which those decisions are made is the product of what has gone before, even when in another sense they may be radical and innovative. The past is deeply embedded in the present. The choices available today are heavily constrained by the choices made by those who went before us – sometimes a very long time before us. That sometimes makes things complicated which seem as though they should be much simpler, and sometimes means that there is no practical solution even when it seems obvious that one should be possible.

Some of that is technical. There is an old story of how the dimensions of the space shuttle were constrained by the design of Roman chariots.  That is alas discredited, but less extreme examples are all too real. The design of tube trains in the twenty-first century is massively constrained by the decisions made about tunnel diameters in the nineteenth. Of course in theory it would be possible to rebore all the tunnels and replace all the trains – but it seems slightly more likely that we will all finally get personal jetpacks than that will happen.1

This problem is not limited to heavy engineering. In many sectors (banking, air travel and government come to mind) even the most apparently modern of systems may rest on foundations going back decades. Nor are the limitations the past imposes on the present necessarily as obvious as the diameter of a tunnel. Charles Stross sums up the broader issue with examples ranging from which side of the road we drive on, through weaknesses in computer languages, to which drugs are made illegal, and makes the critical point that

Part of the problem is that we build rafts of infrastructure on top of existing design decisions. Which means that fixing a bad decision requires the abandonment of lots of stuff that depends on it.

In all those cases, it is pretty clear that we are dealing with constraints and that those constraints do in fact constrain. Providing for potential future change can be expensive in the real world of heavy engineering, and it is understandable that not much of it is done.

Single carriageway road crossed by bridge with spans for two carriageways.

There are two very obvious reasons for that.  The first is that building things for which there is no immediate need costs immediate money but provides no immediate benefits. The second is that there can generally be no guarantee that what is provided for will turn out to be what is needed. Parts of the pre-war German Autobahn network were built as single carriageway roads, but with bridges and other infrastructure ready for a second carriageway. East of the iron curtain, those second carriageways were a long time coming, and driving along those shadowy half motorways remained a faintly surreal experience decades later. The road in the picture above was finally upgraded just a few years ago – but the original carriageway was demolished, not reused.

It should be easier where there is no requirement to dig holes or pour concrete, but the basic difficulties are similar in heavy computing to those in heavy engineering: you can’t easily take account of future technological developments, and once you have built it, it’s difficult and expensive to move. Even if system architects in the 60s and 70s had understood and extrapolated Moore’s law for thirty years, that would have done nothing to change the immediate costs of memory, storage and processing they faced, and the practical consequences would have been non-existent.  That’s less true now in some important ways, but complex established systems are still hard to change. As so often, it may well be clear that there is a better alternative, but very unclear how to get there from here. The principle of designing for future flexibility is largely accepted, even if the practical obstacles are substantial. And even though the new stuff may be easier, the problem of the installed base has certainly not gone away.

And if we take all this up a level again, it becomes an issue for social and organisational change. Cultures, products and processes can atrophy just as surely as engineering solutions.

Most big companies deal with the issue, sooner or later, by going bust or being taken over. Those which don’t can end up in a very different business from the one they started in – it’s been a while since Sony had rice cookers at the centre of its product range.

Governments are not immune to this either, though the stability of governments and governmental systems obviously varies enormously too. But perhaps uniquely in government, there is a strong body of opinion that making design decisions which constrain adaptability to future change is a good thing not a bad thing. The US constitution is a particularly striking example of this effect: its continuity and consistency have taken it through a form of transmutation, where constitutional law becomes increasingly akin to scriptural exegesis. It is for most practical purposes unchangeable: all political decision making has to be built on top of design decisions made over two hundred years ago. A striking illustration of both the short term and the long term stability of the constitution comes from the fact that the most recent amendment went into force over twenty years ago, in 1992 – having been submitted to the states for ratification in 1789. To put it mildly, none of that is seen as a weakness of the US political system by those subject to it: there is no clamour of which I am aware for a new constitutional settlement.2

The point here is not whether the specific provisions of that or any other constitution are good or bad, nor indeed whether having a formal written constitution in the first place is itself a good or bad thing. It is whether constitutions – or anything else – should be designed to constrain the choices of future generations to decide matters. I am not against the idea of constitutions – in the UK context, I quite like the idea of a Constitutional Consolidation Act – or against the idea that they should not be casually changed. But I am not persuaded that I know more about the situation or needs of people fifty or a hundred years in the future than those people will know at that time.

In practice, few constitutions enjoy either the formal continuity of the US system or the informal continuous accretion of the UK approach. The number of countries without a radical constitutional discontinuity over the last century or two is pretty small, and the phoenix approach to constitutional change, of letting the old one burn up and creating a new one from the ashes is probably the most common way of doing it.  But systems so brittle that you can only change them by having a revolution are hardly ideal. My simple solution to the problem of over rigid constitutions is to time limit them. Fifty years sounds about right to me – but of course each constitution would need to contain the conditions for its expiry, since there is no more certainty about the longevity of that approach than of the underlying constitution itself.

That’s not going to happen, of course. In principle forcing the system to refresh itself would allow small issues to be identified and addressed before they got large enough to threaten the whole system, but this is classic innovator’s dilemma territory, so we can be pretty sure that those threatened by change would fail to see the need for it and would have the power to obstruct it, applying what Kevin Kelly has called the Shirky principle:

Institutions will try to preserve the problem to which they are the solution.

So the challenge for designers of roads, railways, constitutions and IT systems remains. Current needs must be met. Future needs must be anticipated, in the certain knowledge that our understanding of what they are decays progressively with time. And above all, the fact that there will be future needs which cannot be anticipated must be anticipated.

Tube picture by Ian Rory licensed under Creative Commons. Berlinka picture from

  1. Update September 2015: I have recently come across a long essay by Mike Horne arguing that looking far enough into the future, enlargement of tube tunnels will become essential and that the scale and complexity of the works involved mean that planning should start now. I remain unsure that jetpacks are not more likely.
  2. Update May 2014:  Eric Posner has since written on this point in much more detail, concluding that the most needed amendment to the US constitution is to make it easier to make constitutional amendments – which is probably impossible without already having done it.


  1. On the railway element, there is often a confusion between track gauge and “loading gauge”. While the track gauge is the same for the London Underground as for other railways in Britain, the loading gauge varies. And the main lines in the USA and the rest of Europe have a larger loading gauge than most of the lines in Britain. Which is why locomotives built for use in Europe by the US military were built to the British loading gauge.

    The US loading gauge was a factor in the design of what became the standard shipping container, and that rippled through into the British transport industry, with only a few rail lines able to handle containers. Some features just had to be moved, such as signal posts on the inside of a curve, but containerisation in the 1960s was an expensive problem for rail transport, and was one of several reasons why road haulage took over.

    One early feature of the US system was the ability to carry loaded motor trucks on trains. There was not enough clearance on a British railway. Similarly, railway loading gauge set a limit on the design of tanks, while the steering of tanks set a limit on the length/width ratio. During WW2, Britain was forced to move from rail wagons to specialised road haulage.

Comments are closed.