Longevity, Repairability, and the Ethics of Manufacture

Last Updated: 2025-02-24 00:00:00 -0600

It’s kind of held as a truism, at least in the west, that we become more cynical as we age. That’s certainly been my experience, though of late, I’d prefer to fight that impulse. With the ongoing decay of my personal mobile devices, and microsoft pushing people off of their Windows 10 installs, and Nintendo making noises about a Switch 2, I’m starting to notice that one area where I’m particularly cynical-of-late is in new device releases. I find it hard to get truly excited when an old favourite studio announces a new game, or a manufacturer announces next year’s model of phone, or when it’s time to get a new laptop. Part of that is just the reality that I’m an adult now, we’ve had several years of local-record inflation, and I haven’t had a pay scale change to match it. But I think that also runs a little deeper.

There’s a popular - and possibly apocryphal - legend that back in the 1980s, Bill Gates defended a relatively low memory limit on an IBM product, that was sort of obsolete-on-release, with the quip that “640K ought to be enough for anybody”. It’s an attitude that’s been hard to shake out of the developer mindset. Somehow, all these bloated electron apps keep getting built even though I hear people all the time saying things like “A gig of RAM just to run Discord is unacceptably decadent”. Go figure. But in this dichotomy, I think I’ve found the seed for my gadgethound ennui.

These machines get a little more powerful every year, but our software also gets a little less efficient every year, and by and large the gains more or less cancel out. My general rule of thumb for replacing computers for the last decade or so has been “I have hit a repairability gap where replacing this machine wholesale is a near-equivalent to the cost of the parts I need”. The last computer I retired outright needed a motherboard and a processor swapped at once because the socket type used by the processor I had wasn’t used on any motherboard then in manufacture. And in general, PCs seems to be the one area where an upgrade genuinely is an upgrade, but only to a point. I’m running modern AAA gaming releases at pretty close to their maximum fidelity - admittedly to the limit of a pretty old display - on a computer that was old when I got it six years ago. For the most part, if I replace a device I don’t get better performance. My phonecalls don’t get cleaner audio when I buy a new phone, my games don’t run better when I get a new computer, and lately whenever you get a new games console you have to keep the old one around just to enjoy your old library. (Digital Rights Management and Media Libraries are a different topic and probably good fodder for another of these articles somewhere down the road, especially given recent events.)

In short, the net effect becomes that I don’t bother replacing anything until I don’t have a choice. I’m not getting the latest Google Pixel 42069XL because I give a good goddamn it’s got 43 extra gregapixels and a sultry AI assistant that can do CSI-style photo manipulation based on voice commands - I’m buying it because my old one wasn’t built to last and I need a goddamn phone or I don’t have a job. To whit: who gives a damn about a new release?

We need to have a talk about longevity and repairability design in the tech space.

E-Waste: A Miserable Pile of Crimes

I think if you live in North America, and probably all of “the west” at this point, you’ve been exposed to the concept of e-waste. It’s this idea that electronics are a special kind of garbage that requires special handling. To a point, this is actually true, for values of true: electronic devices, especially the kind that are computers, contain all sorts of components that ideally need special handling. Lithium Ion and Lithium Polymer rechargable batteries are famously combustable in new and exotic ways. There are rare metals in semiconductor components that are hard to extract from the wild and are better off not just in a landfill. There’s just one problem: the solutions solution’t.

Recycling: The Wouldn’t-It-Be-Nice Industry

I’m going to ignore for the moment that in most municipalities in Canada (and probably extendable throughout at least North America, if not the whole world) have miserably inaccessible systems for handling e-Waste. I’m not ignoring it because the problem is unsolvable; it isn’t, and we could discuss that in the future. I’m also not ignoring it because the fact that e-Waste collection objectively sucks would undermine the thesis; even perfect collection doesn’t make my broader point here wrong. Instead, I’m blowing past the reality that most e-waste winds up in landfill rather than the proper collection streams to highlight something much deeper and with broader-reaching consequences: Recycling Doesn’t Actually Work.

Yes, it is possible to extract some materials from old products and return them to a sort of “raw stock” form that can be used for other products. Certain metals are valuable enough to be deemed “worth” extraction from waste - gold is the most popular example here, but recycling cobalt is crucial in terms of lowering the world’s net suffering. Glass can by-and-large be reused (though some applications - like good optics - require “fresh” glass made from pure silica sands). Plastics, though? Astoundingly diminishing returns.

The very short version of my argument here is that recycling is the bilge pump of the ship of Industry. Yes, you will always have a waste stream, and you need to remove that waste before it sinks you. But if you’re taking on water too quickly, the best-designed bilge system in the world is still going to get overwhelmed, and you damn well better know where the exits are when that happens. Even with better processing standards, even with the mythological 100% waste stream capture, there’s just too much smart nonsense entering the waste stream. The real magic trick would be to reduce that figure, and that, unfortunately, becomes a problem of economics. Don’t worry, we’ll get there.

Cobalt Mining, or “What if we did Cibola Burn in real life?”

Lithium Ion batteries actually contain two metals. Lithium is just the anode. There’s another metal that is extremely rare, thereby valuable, and is absolutely vital for the modern formulation of lithium batteries, which are in pretty much everything because they’re the best rechargable chemical battery solution we’ve got, at least until someone out there really figures out the supercapacitor. This of course means that those batteries are in damn near everything that we’ve cursed into thinking: phones, laptops, tablets, smart watches, and that bullshit “AI Lapel Pin” thing. The original version of PETI’s design spec called for an integrated, chargable lithium ion battery pack. Cobalt is also used as an adjunct metal in a whole bunch of steel alloys that are used for constructing this stuff, too. There’s just one problem with our friend cobalt: its extractive processes are human rights travesties.

By far the largest gobal supplier of cobalt since the early 2000s is the Democratic Republic of the Congo, and if it wasn’t for a series of complications we’re definitely about to get into, there would be no problem with that. Hell, it would be a major victory for African Economics, right? A free African state and free African enterprise is the global supplier of one of the world’s more critical industrial inputs. Except, of course, for all those pesky complications.

There’s the usual objections to be made to most mining, of course: safety is horrendously under-regulated, and mining polutants are a hazard to indigenous communities and local wildlife, same as it ever was. And of course, there are forms of colonialism that aren’t settler colonialism: the vast majority of Congolese cobalt is contractually obligated to enter the wider world market through a small number of Chinese-owned mineral corporations, like Congo DongFang International Mining, or Zhejiang Galico Cobalt & Nickle Minerals of China. Back in 2016, it was estimated that something like 10% of the entire global cobalt supply flowed through Chinese corporations. This is taking a country that could basically be the Saudi Arabia of the rechargable electronics age (including for electric vehicles) and making sure it stays poor. So, you know: standard colonialism.

Surprising nobody, it actually gets worse. Because of the under-regulation I mentioned above, it’s sort of well-known at this point that Congolese artisinal mining for cobalt (which makes up about 40% of their production by some estimates) employs child and forced labour. Both are obviously pretty significant human rights abuses, and both are kept as far from mind as possible for all but a few activist types in the global north. The situation with these abuses are so well known that Apple, who normally I’d tear to shreds in an article about this sort of thing, actually announced in 2017 that they wouldn’t buy ore from the suppliers that work with these mines, and began to use only suppliers that are verified to meet its workplace standards. This of course doesn’t mean much, because Apple thought Foxxconn was okay. In 2019, a human rights watchdog NGO called “International Rights Advocates” file’d suit against Apple, Tesla, Dell, Microsoft, and Alphabet for, in their words, “knowingly benefiting from and aiding and abetting the cruel and brutal use of young children” in mining cobalt. While in 2024 the courts that saw this case did hold the suppliers responsible, they managed to find an excuse to aquit the US-held Tech companies (again, same as it ever was). As a small nod to this, in 2023, Apple pledged to only use recycled cobalt from now on.

We need cobalt. Congolese workers deserve the same human rights and safety regulations as the rest of us, and they deserve to capitalize on their country’s natural wealth of cobalt the same way any other country would. Until international economic pressure or internal labour pressure forces that situation, though, we have other choices. The second-largest producer of cobalt in the world makes 20% of global production themselves. It’s Indonesia, and the cobalt is a side product of their mainstay production of nickle. Canada, where I live, thinks it has some pretty significant reserves, but we’re only fourth in the world. Third is Cuba. We like to give Elonmitri de Melonsku no end of hard times because his family’s wealth was in apartheid emeralds, but nobody cares about enslaved black kids losing limbs to make sure they get their new gizmos. That’d be inconvenient personally.

But!, I hear you exclaim: if everyone was paid fairly for their work cobalt mining and had first-world safety and labour regulations to protect them, how would anybody afford a new phone every two years? There’s a very easy solution to that: they probably wouldn’t, and that’s perfectly fine. (Though, we’re also looking at the possibility of alternative chemistries for these batteries, like using lithium iron phosphate instead of lithium cobalt oxide. If that worked well enough we’d basically take the problem of cobalt mining out at the knees.)

After all, the only real reason to buy new hardware every N years when they aren’t materially improving is that they also stop working about every N years, for a variety of problems. So, if we can solve that problem, we can indirectly solve the cobalt and recycling problems because now we need less of the unethically-mined supermetal and we generate less of the stuff that we aren’t actually bothering to recycle anyway. So… let’s try to wrap our heads around that.

Unsolvable Problems in Smartphone Design, or “You Can’t Fix Stuff Anymore”

It should probably surprise nobody at this point that I am a right to repair advocate, but I think that needs to go deeper than policy. Repair is fundamentally an engineering problem. That’s not to say the need for repair is somehow something you can engineer out of a design: that’s impossible. There’s a few laws of physics that will come along and club you over the head if you try.

In electronics (and software, but that’s another matter) there’s a well-known engineering concept called Mean Time To Failure (MTTF). It’s sort of the LD50 for tachyons - the amount of operating time at which roughly half of all the devices of that design are in a failed state. It’s a really common metric on hard disk drives, but it applies to literally everything. There are other metrics that cover this sort of thing too, but that’s not really important. What’s important, fundamentally, is that we know from square one that whenever you manufacture anything, it will eventually fail. If you’re a good engineer who gives a damn about product quality, you plan for that eventuality, and plan for repairability. I’m not sure that’s what’s haappening here.

You are not meant to fix this.

Sometimes, I give too much lattitude to the idea that for a lot of product designers, design for aesthetic is a higher priority than design design for service. That’s the excuse usually given for why just about every phone on the market is more or less glued together. I no longer feel so charitable, so I’m looking at that excuse a bit more critically. I don’t think companies consider repair profitable, and I think fiscal responsibility is a sort of neurological condition that causes you to hyperfixate on the profit motive. In general anything that increases the lifespan of a device sort of lowers the overall profitability of the product line because it impacts sales. I’ve said more than once that if I could keep my phone maintained I would probably never replace it. Never replacing it was the goal of that PC I was forced to replace a few years ago by not being able to afford to repair it. It’s the goal of the small fleet of devices I have now. In an industry where innovation is largely illusory and people are starting to find their They Live glasses, device failure is increasingly the main upgrade driver for all but the most terminal of gadget-hounds.

It is possible - even easy - to design devices for repair. It requires some concessions. Sometimes in overall product size or manufacturing complexity. Sometimes in having visible fastners spoil your highly destructable “pane of glass” finish’s aesthetic. You don’t even have to compromise on product utility. Most of the major manufacturers of laptop computers have at least two product lines - one for consumer, and one for enterprise. These lines often have very similar specifications in terms of things that would impact their performance. The major difference? The enterprise computers have some effort put into their design to make life easier for the service techs that have to keep them alive. I used to physically repair computers for a living, and while this is anecdotal, I can state pretty emphatically that for the major manufactures - Dell, Lenovo, and HP - their enterprise lines are all easier to repair and maintain than their consumer or even prosumer models. This is a known thing. The OBD standard exists in automotive to make diagnosing hardware faults in the automobile easier for the person who has to repair them.

We could be building repairable devices, but we aren’t, because it’s not long-term profitable. Every 10 year device you construct creates a client that isn’t likely to make another major purchase from you for 10 years. You’d RATHER sell a $1300 handset than a $30 Li-Ion battery pack for it.

The Black Box Problem

The existence of Free Open Source Software (FOSS) and Open Source Hardware (OSHW) standards notwithstanding, the overwhelming majority of everything you can buy is fundamentally a black box. We are now long past the era when you could reasonably seek to obtain a service manual for a given device or appliance. It is actively expected that you rely on the benificence of the manufacturer with regard to troubleshooting and repair documentation. If you do manage to get a device open, effort is often expended on both making that difficult and obscuring the parts numbers normally printed on given components, up to and including doing things like applying resin “blobs” overtop of component sets to thwart replacement or observation. This is also true in software, where documentation for internals is a pipe dream and documentation for APIs is often barely-useful even for experienced developers. This design philosophy makes it harder to diagnose and correct an issue in a device even if you know what you are doing.

Warranty, and other lies we tell ourselves.

The good news is, the average person shouldn’t have to repair their devices anyway. The practice of offering a product waranty has become more or less ubiquitous in North America and, I think, most other places where the practice is legal. For the overwhelming majority of technical devices, these waranties fall into two categories: a) Effectively useless in that they only cover manufacturing defects, meaning basically any failure that can be called user error will be called user error and the user is SOL. Go buy a new gadget. b) Paid-extra extended warranties that would also cover user-caused damage, but which include provisios that the warranty provider may just “replace” the device instead of actually repairing it.

These warranties have a lot of drawbacks. First of all, they are almost always for a fixed term, which doesn’t exactly create a profit-disincentive to build a product with an MTTF longer than the warranty period, but it certainly doesn’t incentivise the opposite. Both the useless and pseuoduseful kind of warranty also usually include riders that automatically void the warranty if the device has been tampered with in any way. A fair amount of engineering and product design goes into using tamper-evident seals to enforce these conditions, and these seal technologies are… suboptimal. I’ve seen moisture-sensitive stickers turn to their “has been immersed in water” colour from being operated for extended periods in humid conditions, and I’ve seen warranty seals that are supposed to only be broken to access a fastener simply fall off if the device gets warm enough. This condition is the real mechanism by which warranties economically disincentivize repair. I’ve heard most people in my life express at one point or another that they’re better off filing the warranty claim and going through the hassle of outright replacing an entire device than trying to repair a minor issue themselves on the grounds that if they got the repair wrong, now they have a broken thing and no good mechanism to replace it. In an economic regime where some of these devices are qualifying equipment to be able to do our jobs but aren’t provided by our employers, we are seriously disencentivized to mess with the warranty.

Repair is Not Trivial

A common lie that a lot of us tell ourselves, meaning both casual supporters of Right to Repair and folks who work intimiately with technology, has to do with skillsets. We both tend to think that “the average person” (which we really just conceive of as ourselves with the fashion and serial numbers filed off) is both more technically inclined and technically educated than the actual societal average. We also tend to tell ourselves we’re more skilled than we are, because the first rule of Dunning-Kreuger Club is that you don’t think about your role in Dunning-Kreuger Club.

The thing is: electronics repair is not-trivial. We’re not in the 1980s anymore: the age of repairing a stone-dead commodore 64 with four inches of bodgewire, a handful of Nihon Chemicon capactiors, and one or two easily-swapped DIP ICs is more or less done. Even eliding the truly miniaturized stuff like cell phones, a modern PC is just more machine than one engineer could hold in their head at once. Solving the black box problem with some kind of mandate or incentive to provide good documentation wouldn’t actually alleviate this issue. Even if you did fully comprehend the electronics of your smart phone, the manufacturing tolerances and component sizes put the kind of solder rework it would take to repair a true hardware failure in a device firmly into the realm of “specialist labour requiring a small business’ worth of specialist equipment”. That kind of repair is an ideal that I don’t think ever truly existed on contemporary equipment in the digital age, except, again, maybe in the very early days of the microcomputer.

There’s good news on that score: most repairs aren’t really like that. I think that’s a good ideal to maybe chase for the very long term, but there’s shorter-term solutions, implementable now, that work great. With the exception of two phones that I couldn’t use any more because specific broadcast technologies, every cellphone I’ve owned has died for one of two reasons:

Some damned fool (who shall remain nameless) has physically damaged the screen, or;
The battery did what all batteries do and lost so much capacity as to make the device impractical to use.

And the great news on that score is that modulized parts of devices are super repairable, compared to having to get into the guts and do individual component rework. Back in the day, phone back covers just came off and you could simply swap out the battery like the batteries on your TV remote. Extra battery packs were actually a really common accessory sale. Screens are usually pretty easy to replace too, provided your manufacturer hasn’t done something really sketchy, like that time [Apple got caught using read-only properties of device components to stop replacement screens from being used on iphones]. Why hamstring repair on purpose? Well, that’s easy: apple doesn’t make any money when your local corner electronics repair fixes your phone, because they don’t actually sell parts.

Most things can be modulized. A big part of the reason I still favour the desktop PC over even a laptop is the sort of infinite-replaceability that comes with the ATX standard. In general, if any one component of a PC fails, it can be replaced, without having to scrap the whole device. Laptops are less like this, and phones and tablets even less so. But there’s no good reason for that, other than design-for-repair takes work.

Good News Here:

There are exceptions. I’ve seen pitches for products that I can’t endorse because I’ve never used them, but where I like the cut of their jib. There’s a company out there called Fairphone whose most recent model, the “Fairphone Five”, has a trivially removable back plate that allows you to swap out the battery, screen, USB-C port, front or back camera, or loudspeaker as swappable modules. No specialty knowledge or tools required, just follow the documented tutorial. The whole phone is still a reasonable size comparable to, say, my Google Pixel 4 XL, including the weight. If their marketing materials are to be believed they’ve addressed a lot of my concerns expressed above in labour. It’s warantied for five years though, and they’ve promised OS support into the 2030s. It’s about as sustainable as that industry gets.

The only real downside is that no Canadian carrier carries their device natively. So I can’t take advantage of any kind of the usual financing structures to get one. Price wise they’re not incomparable to mainstream phones which is impressive considering the economic forces involved, but there’s a big difference between a ~$1800 purchase hidden as two years worth of $10 surcharges on my phone bill and being able to come up with €549.00 on-the-fly.

One Step Further: Design for Longevity

I opened the section of this article on repair with the assertion that the need for repair is inevitable, and that statement is true. There’s an old saying among the wise: schedule maintenance, or your equipment will schedule it for you. The existance of a MTTF means that that time is set in stone. You must suffer equipment failure (or forestall it with maintainance where possible). You do get get to influence, at the design level, how near or far in the future that MTTF is.

The thing is, with a lot of electronics, this isn’t as much of a limiting factor as it seems like it might be. Nothing I’m saying here necessarily suggests that companies are really deliberately setting out to make their mean time to failure low on purpose. They’re properly mitigating for easy causes of failure: no amount of engineering will make their batteries wear out slower, short of new chemisty, so the solution is to make batteries easy to replace. But for the most part, electronics are basically solid-state devices now. Laptops have hinges, keyboards have keyswitches, and mechanical connectors have springs and tolerances subject to wear, sure, and solid-state memory actually can wear out too, but those are being pushed near their physical limits. We’d need new physics to make NAND memory cells last longer.

Indeed, the one area where design for longevity could be improved in computing is at the software level. The hardware is fine.

We’ve nearly expended Moore’s Law, and year-on-year the returns for hardware improvements have been more or less shrinking for general computing. I wasn’t exaggerating when I said I haven’t really noticed an upgrade in performance between machine upgrades. Sure, I notice a few things when I change something major - I became an SSD convert the first time I had to boot up my computer off of an SSD and saw just how fast you can raise a computer from cold glass to working system session that way. But day on day, the same problems remain: more than 20 browser tabs is asking too much. God help you if you want to be in your social chat programs AND run VCV Rack.

If computers keep getting better every time we replace them, what’s the deadlock? Software. It’s getting a little less performant every time we update it, and it just so happens the tolerance peak on that math means that the performance degredation in general aggregate more or less follows Moore’s Law too. In effect, we break even (or, given that I’m complaining about it, even come out behind.)

Now, Longevity and Sustainability in software design could be a whole rant into and of itself. Trying to compare Windows for Workgroups 3.1 to Windows 11 is a little like trying to compare a bottle rocket to the SpaceX Crew Dragon: ignoring for the moment how badly designed both Windows 11 and the Crew Dragon are, they also are just so much more complex under the hood than their counter-examples. There’s a lot to unpack there.

Because you know what the other killer of a phone is? Say you don’t need a mountain of battery life and you’re gentle on your gear, but your phone’s like 6 years old. Your manufacturer stopped shipping OS patches three years ago and now your app manufacturers won’t let you install the latest versions of your favourite gacha game or your banking app because the newest versions aren’t compatible with the old OS, or they have security concerns.

In a way that’s more real than a lot of us tend to like to think, software is as much a part of the machine as the hardware is, when working with digital devices. And we’re really, really bad at designing software for longevity. Hilariously bad. I sometimes like to joke that if Software Engineers were really Engineers, none of them would be employed, or that sooner or later the International Association of Professional Sofware Engineers will induct new members by giving them rings forged from the burned-out metals of the ruined AWS US-East-1 Datacentre. Our willfull indulgence of feature creep and the industry-wide insistence on high-velocity releases, forever chasing the New Shiny so that we can be the bower-bird that gets the customer that contract term has basically killed software engineering. It’s just dead like absorbing a few hundred Greys all at once: we’re still moving, and nobody’s told the brain yet.

The lack of sustainability in software design is fodder for plenty of future thought and writing, so I won’t belabour the point here.

How do we fix this?

Like all problems, the solutions are obvious and straightforward and the implementation is a spaghetti ball of “who the heck knows”. If these problems were purely engineering, they could be fixed overnight with new standards of practice. The problem is, these problems are economic. We chase Moore’s Truism and the new shiny and all the other things that are driving down Mean Time To Failure because we have fiduciary responsibility and the Number Must Go Up. You could kill this problem overnight by eliminating the profit motive, but doing that would require creating for yourself several hundred new problems. You could probably solve this problem in regulation, but the problem is that regulation is by definition written by non-experts. In much the same way that middle-managers do a different kind of work than line engineers, legislators do a different kind of work than… line engineers. By definition.

In general though, there’s at least targets:

Eliminate legal cover for warranty provisions or licensing agreements that prevent the user from repairing (or allowing third parties to repair) their equipment.
As engineers, focus on design decisions that eliminate waste by maximizing the survivability or repairability of hardware and software.
Where possible, as engineers, favour providing services to organizations that align with higher standards of sustainability, device longevity, and repairability.
Provide regulatory or economic incentives that drive MTTF as high as possible and encourage (or require) manufactures to make replacement components available.
Where possible, as consumers, limit consumption. When consumption is unavoidable, favour manufacturers that support pushing the repairability envelope as high as possible, and who support open standards.

These are just a few ideas, from one lonely sort-of-Engineer stuck in the innovation backwater that is the software industry. But, I think, at least these five should be found mostly-agreeable by the majority of you.