> Put another way, if 'software is eating the world', then if it's free software the cost of goods and services should drop dramatically.

Software cost generally isn't a "dramatic" part of cost, wherever you look. Because of zero marginal cost, the price of widely-used software approaches zero naturally. Only highly specialized software commands high prices, those reflect the cost of developing or replacing it, as well as the value that it is providing.

Is that true? I can buy a blank laptop for £215 (https://www.pcspecialist.co.uk/notebooks/gemini-IV/) and then install Debian for free. The same laptop with Microsoft Windows / Office costs £352. So I'd say that in this case at least, software is a dramatic part of the cost.

You were talking about the "cost of goods and services", by which I assumed you meant COGS as a factor of running a business.

Let's say you put a minimum wage worker to some task on that laptop. Over the amortization period (3 years) of that laptop, that extra cost of the Windows license amounts to maybe 0.1% of the cost of that worker, assuming they use nothing but that laptop to perform the work.

Now, let's say that the worker needs to learn how to use Linux and it costs you, say, 20 hours in opportunity cost to get them up to speed. Linux is now more expensive than Windows.

Okay, let's go with your figure of 0.1% of the cost of the worker. My point is that if society were to switch to Open Source software that 0.1% is compounded because all the other goods and services that a business has to buy become cheaper. This then means that they can lower their prices while still making the same profit. Which in turn lowers the cost of inputs to other businesses in a virtuous cycle.

So using your analysis of the single worker, which I agree with, it actually makes a very big difference when applied to all workers.

To address your point about the training costs of moving from proprietary to Open Source, again this disappears if the move happens at scale. It disappears because every worker would have learnt Linux at their previous job (or at school / college etc.) and so in fact the cheapest thing becomes to use Linux.

What your argument demonstrates is the network effects of proprietary software, whereby if you're the first to use Open Source software it's a disadvantage, but if everyone used it, it would be an overall advantage to society.

The question then becomes, how do we get from here (often proprietary) to there (mostly Open Source)?

> So using your analysis of the single worker, which I agree with, it actually makes a very big difference when applied to all workers.

It doesn't, because it's still just 0.1%. That may be billions upon billions of dollars in aggregate, but per unit sold, it's still just 0.1%.

So, if your question is "how much cheaper could goods become by switching all production to free software?", then answer would be 0.1%, if we assume that 0.1% is the average share of software licenses in COGS.

In practice, software licenses averaged over all industries might be 1% or higher. Again, that might be trillions of dollars, but it's still just 1% per unit sold. Also, that money isn't evaporating, it's mostly going into the pockets of people who build that software.

Sure, there's a profit margin in software development, maybe 10% on average. So, if you eliminated that profit, you'd save 10% of 1%. You'd still have to pay for that development, even if it is free software. If you eliminate the profit motive, what governs development? Some form of bureaucracy. We already know bureaucracies are inefficient. So you would probably end up paying more to develop "free" software.

From an efficiency perspective, free software is strictly unnecessary, because the marginal cost of software is zero, the cost of developing it is a constant, and therefore its price naturally converges to zero as its development is amortized.

Let's say OpenSSL was proprietary and cost $1000. That would add $1000 to the cost of developing anything that uses it, until a cheaper competitor comes around. How expensive is it to implement SSL? Not that expensive. How many people use SSL? A lot! Do the math, make the business case, and you would probably find that you can turn a profit selling a competing implementation at only 10$. Why is this not happening? Well, because somebody already made OpenSSL free and it's "good enough".

It is a misunderstanding of market forces to assume the compounding effect of free software licenses would be greater than the compounding effect of cheap software licenses that the market produces all by itself.