I used to work as a civil service engineer for the Navy in technology development for future weapons. The brass would often complain about how expensive fully-i

I used to work as a civil service engineer for the Navy in technology development for future weapons. The brass would often complain about how expensive fully-integrated solutions were. We investigated and found that the main cause was vendor lock-in to buy the all up rounds. We couldn’t compete for lowest cost on the power cards, or signal processing cards, rocket motors, etc., because the government didn’t have data rights to the design in order to hold down development costs. (Side note: as a contracting necessity, we’d also put the whole purchase on one huge contract to be managed by the so-called prime contractor and negotiate a profit fee explicitly in the contract itself. The total cost of the prime contract included “pass through” funding for subcontracts that the prime contractor would handle. In some cases the prime would let a subcontract to a vendor who would contract _back_ to a subsidiary of the prime, and we’d pay guaranteed profit margin explicitly on the prime and the subcontract amounts, so we’d pay a profit markup to the prime on the profit they would make from their subsidiary. And folks around here gripe about the 30% app “tax”.) Anyway, vendor lock-in means we can’t compete subcomponents and our long term maintenance costs are really high because we always have to go back to the vendor for service. (Sounding familiar…?) The brass started dreaming up “modular weapons systems” where we maybe develop things like universal power modules, guidance modules, and other components, then we could compete for low cost production and long-term maintenance and achieve utopia. I spent quite a while on this, and what we found is that it wouldn’t work (at least not for weapons). Universal components aren’t well-suited to different platforms (3” rocket vs 14” missile, etc.), the modular design adds weight, reduces efficiency for the electronics systems, increases part count to make them all interoperable, and so on. The engineering trades that you have to make for modular designs are incompatible with optimizing performance as measured by the end user. Ironically perhaps, one of the major case studies that helped argue that point was based around Apple products. Which is perhaps a long way to say that most consumers don’t seem to care about modular computers, and in the worst case many of the changes that modularity would require are counterproductive to the things consumers _do_ seem to care about. It’s one thing to argue that we should use governmental regulation to push negative externalities back onto firms so that they reconsider the design and engineering constraints they prioritize, but it’s not like firms are being irrational or irresponsible by giving people what they want—some combination of smaller, lighter, faster, longer battery life, and cheaper devices. There would be some negative impact on some (or all) of those characteristics if Apple, Samsung, or anyone else switched to fully modular designs—and I think most consumers would choose the cheaper, faster, smaller, lighter, or longer battery life models than to buy a modular one either on principle or because they expect to repair or upgrade it for significantly longer than they expect to own them today.