Why It Is Important That Software Projects Fail (2008)
berglas.org
berglas.org
I love this conceptualization of what software is for. But I have the complete opposite moral interpretation.
Unstated is the presumption that bureaucratic complexity is bad. I think increasing it is the goal of society, and ethics constantly demand us to increase it.
For example, it’s easy for banks to simply reject loan applicants who were convicted of felonies. There’s no easy upside: given existing loan approval and repayment processes there’s no clear way to give those loans profitably.
There is a huge social cost to that group lacking access to credit though. It increases crime and violence.
And there is probably a long term financial cost too: some felons will reliably make their payments. And still more COULD reliably make payments with the right support.
So how does a bank separate out those groups and provide that support?
Increase bureaucratic complexity!
I am not necessarily in favor of locating that complexity in a state government. As an anarchist I would want to locate it as close to the affected community as possible....
But to me the dream of software is nearly infinite bureaucratic complexity that affords all of us greater freedom, and health.
For example a lot of decisions in Anarchist communities will be driven by arriving at consensus of all concerned by the decision, instead of a "mob rule of the majority" like in Democratic voting process.
The way you arrive at the decision is a bureaucratic process if you will, that the community has agreed upon.
That's my take on it
To quote a relevant passage: "Anarchism is a process whereby authority and domination is being replaced with non-hierarchical, horizontal structures, with voluntary associations between human beings."
If bureaucratic complexity is the complexity of the structures that organize our society, its existence is entirely consistent with an anarchist society. The structures would no longer be hierarchal, but anarchism makes no claim that organization (or the complexity associated with it) shouldn't exist.
Comprehensive bureaucracy can be good but also often fails to efficiently handle problems and gives advantages to specialists.
The result can be something like this:
https://en.wikipedia.org/wiki/CumEx-Files
> So how does a bank separate out those groups and provide that support?
Isn't it a bureaucracy that provided the information about felons in the first place? Not saying that interest in information is invalid, I am just doubtful about the solution.
Not necessarily. The main difference here is that most bridge projects are not private investments and so cost overruns are less likely to cause complete failure, but the cost overruns can still be massive:
https://www.mercurynews.com/2018/05/03/letter-bay-bridge-cos...
The article he cited, Parkinson 1955 (originally in the Economist), on the relationship of british naval bureaucracy to work, is even more interesting imo! http://www.berglas.org/Articles/parkinsons_law.pdf . It hypothesizes a growth equation for bureaucracy where the only exogenous factor is the underlying population. I'd be curious whether more rigorous research says anything about bureaucratic growth, as we're stretching further towards an era where minimal % of the population has to do any 'blue collar' labor. Maybe in the field of institutional economics?
Having observed additional layers of bureaucracy develop over time, I have noticed the additional factor of attempting to get blood from a stone. When a the critical metric (eg revenue growth) falls, a perverse incentive of increasing control is pursued after some short gains. Following the pereto curve, this too is less effective over time.
This is an excellent expression. I have observed this sort of thing as well, especially with regards to deadlines that get pushed back.
The gig economy may cut out "a lot of bureaucracy", but it's replacing it with other forms of pointless work.
I don't think this is more true anywhere than in software. How many AngularJS developers does it take to change a light bulb? Answer: Never mind, we're rewriting the frontend in React.
Surely it's more, now a bunch of "contractors" have to do loads of company admin off the clock.
It's just a dodge, shifting bureaucracy and costs from a company onto an individual who's being forced to masquerade as a company. And possibly a flash in the pan, it is certainly possible the gig economy will be an illegal working practice in 20 years time in socially progressive countries.
The p2p gig economy does not drive capital growth or replace labour.
The thing about the automation of bureaucracy is that what it being produced doesn't a value but rather has a "redistribution effect", so more tax codes produce at best a more fine-tuned rearrangement of wealth according to the ideas of legislators. So increased productivity is meaningless.
One thing I'd add is that there areas that produce useful items but where the insertion of automation has helped much either - health care is an example I find interesting. And certainly there are places where computer-based large-scale automation has resulted in serious increases in productivity (I believe rail roads had this in the 1990s).
So question of where automation tends to fail and where it tends to succeed is a bit open - but it seems like "organizing complex, interlocked human interactions" is one place where failure is common.
In the absence of other hard limits, project failure is the only reality check on software development. Without that, we'd all live in Cockaigne, fried chickens flying into our mouths.