With increased integration, there's also an aspect of the cost of a module being almost independent of its contents, meaning that the cost of making a device is roughly proportional to the number modules it contains. Under those circumstances, it makes sense to minimise the number of modules, the lowest overall device cost being a single integrated unit.
In my mind, the key to long life is freedom to modify the software. It's amazing how far a well designed hardware platform can be pushed with software optimisation over time. I do embedded systems in my day job and our current product does things that weren't even thought of when it's hardware was designed, by virtue of the original design keeping the hardware simple whilst doing as much as possible in software, and current creative use the existing hardware resources.
I think a fixed hardware platform can work well, provided an original design goal is to do as much as possible in software and there is complete freedom to rework the software going forward.
Is there an opening here for a non-manufacturing "open" standards body to emerge? They could publish mechanical specifications for a "click together phone" and a specification for a simple electrical bus/interconnect (use USB for everything?)
One could envisage a series of standard exterior cases, based on a common unit of measure. Each case might contain openings for screen, camera, connectors, etc. according to the type of device. Inside the case would be spaces for internal modules, also based on the common unit of measure, not specific to any one type of module and with interfaces in standard positions. Exterior cases in the series might include:
* Smartphone
* Tablet
* Laptop
* Desktop
* Media player
Such a movement could arise from the Makerspace movement, as it would be a relatively low cost thing to do? One can imagine a website with specification documents and 3D printer files for the various module outlines. People would then be free to experiment on making new components within this ecosystem.
---
Edit: I'm pessimistic that the above would progress past a hobbyist level as the economics favours an integrated approach, unless it spurs a level of competition and innovation that undercuts the premiums being charged by the makers of walled gardens.
Even when Google was still working on Ara, they planned to release the "Endo", or base frame, itself, and then just let vendors sell modules with a similar OEM partnership to how they control Android OEMs today.
It worked in 90's and early 00's, particularly with RAM, hard drives and CD ROM drives. You could walk into a consumer electronic store and buy these drives stand alone, I don't think you can do that anymore. I was frequently upgrading friends and family computers for beer, anyone technical usually had a bunch of old RAM sticks and hard drives lying around for anyone that wanted them and a new computer frequently meant a new motherboard and case where you could add all the component of the old one.
This was a time where hardware upgrades were a huge deal. Going from 8MB to 16MB of RAM was a lot more valuable than going from 8GB to 16GB is today. And there were frequently software releases that warranted hardware upgrades. For my, buying visual studio necessitated I upgrade my 540MB hard drive to a 3GB one for instance.
> After a few years, it becomes a cascade of dependencies: Can't upgrade memory because the old motherboard won't support state of the art DIMMs, the bus used by current cards (ISA, PCI, AGP, PCIx) doesn't match the motherboard, power supply connector has different number of pins, and if you replace the motherboard to allow a newer component to be plugged in then all of the other old stuff needs to be replaced due to the new motherboard.
Yeah, it became like that in the early 00's from memory after a decade of relative stability. That's when the hardware world got too complicated for a simple software dev like me. Although with everything on SATA and PCIx these days I wonder if I could pick it up again.
Stores like Fry's Electronics have major component offerings, and even at your Best Buy or Staples today there's generally a limited supply of standard replacement parts for conventional PCs.
With DOS and Win9x one could effectively replace everything underfoot of the OS and it would hardly blink. Yes it would fall back to 640x480 and 16 colors, and demand you install some new drivers, but beyond that you could do just about anything and it would still boot.
Come NT and you would just be greeted with a error screen and a demand that you did a recovery (that would effectively roll back all patches etc).
And these days you have to phone Microsoft if you try anything like it more than a couple times because you tripped their DRM checks.
Yes, one could wander into a solid case of "my grandfathers axe". But the various layers were independent enough that they would come up in some kind of state that you could work with.
Sadly i fear that Linux is heading into the NT realm with the changes to the graphics sub-system that has come in recent years.
Where before higher modes were the realm of X, and things like 3D acceleration was yet another semi-independent layer, these days the kernel may well give you an error if you do not have the right GPU driver compiled in.
(I can't remember exactly which version of Windows this was!)
IME, it never did. Except for techies, hardly anyone opened their computers. I remember seeing some system offered to corporate customers that lacked upgradability (except maybe memory). At first I objected - what a ripoff - but then I realized the truth.
My impression is that PCs got their start with hobbyists and were designed for them, with extra slots and bays, etc., and to this day the designs are still adjusting to the vast majority of end-users. I think the first office suites included relational database authoring tools for everyone - as if most office workers would just whip them up as needed, like word processing documents.
Apple has always seen it differently: You couldn't (can't?) open the cases without special tools.
FYI the Mac Pros before the trashcan had a latch and the side opened just fine for upgrades. Actually even the trashcan can be opened but I kinda get the impression it can only be opened for you to admire it not really for you to upgrade it...
Upgrade memory: https://support.apple.com/en-us/HT205044
Upgrade storage: https://support.apple.com/en-us/HT202881
It did in the early days (the first 10 years). These days, the mobo, CPU, ram, power supply, graphics card all seem to need to be a matched set.
Now that CPU speed increases have leveled off things seem to have stabilized again.
A comparably-modular mobile phone cannot be anything but thick and heavy, unless you would prefer fragile and unusable.
Android phones have NONE of this. Google could force this. Their "Open Handset Alliance" is not at all open. It's a closed lock in system to keep all their manufactures from ever making Amazon phones.
Google has a vested interested in selling more phones. They could easily require standardization. They could ban binary blobs. They could require all manufactures to make their kernel patches and drivers upstreamable and meet Linux kernel standards. But they don't.
You can take a 32-bit x86 Linux distro from today and run it on an old Pentium. You can take a really old 32-bit distro from years ago and run it on a modern machine (of course very little of the hardware will work, but it should boot in legacy mode...okay, okay it might not boot, but if it does boot you will get a console I/O).
It has nothing to do with the phone and has to do with the fact that ARM is not an architecture. It's a SoC sold to tons of manufactures who all connect whatever they want to onto whatever pins they want to. Yes yes, there are things like Device Tree configs, but name a phone that supports them (no seriously, I'm asking. Are there any? Like I would want to buy/experiment with them if they exist).
Every ARM device seems to have a totally different base image and this isn't maintainable. We need a standard device/IO configuration system so that you can install standard AOSP + drivers onto any phone just like we can with Windows:
If that were impossible, Ara would have not come as far as it did, and the Fairphone 2 would not be as modular as it is.
Soldering can get things pretty small, and pretty close together. Home soldering is possible, but requires some tools, and a steady hand. if you want people to solder, you'll probably need to make the connections a little bigger and further apart than what you would need for a nice smt with mask in a shop. Soldering is is also kind of hard to reverse.
Any mechanical fastener is easy to undo, but much more bulky than soldering. i'm thinking little plastic clips, but there are alternatives. in any case, they're going to be bigger.
That leaves maybe some kind of conductive epoxy and a solvent to remove it. that should be pretty comparable to soldering, but seems like it would have the same problems, you'd want bigger surface mount pads.
mechanical fasteners are the way to go, you don't need any (or many) expensive tools to get it right. but mechanical fasteners are big, and that's space that could go to more battery, bigger screen or whatever. further apart means slower bus speed, and probably other problems.
The other options require special tools. So phone plus $200 or whatever for the tools just isn't worth it. (and making the process super reliable for novices isn't easy)
I don't quite understand this point.
PCs are inherently upgradeable, but it seems like consumers would rather treat them as disposable electronic devices, judging by the success of barely-upgradeable Macs and other laptops.
And I'm not sure I blame them either, it's not like it's easy for a consumer with no techy knowledge to know whether they should upgrade the RAM, or the SSD, or the GPU, and research which components are compatible, best bang for the buck, etc.
You're making out upgradeability to be the reason why PCs won and are super successful, but it seems more because there were a whole bunch of PC-compatible clones that could all run the same software. The hardware/peripheral compatibility between PC clones was never very particularly good until well into the 90s, well after the PC standard had already 'won', and the computers they replaced were also upgradeable.
Disposability is a side-effect of non-upgradeability - consumers don't explicitly want either of those, they just want a good product. And a good product can't be upgradeable.
See Daring Fireball's article on this: http://daringfireball.net/2011/09/new_apple_advantage
The money shot is, "Apple offers far fewer configurations. Thus MacBooks are, to most minds, subjectively better-designed — but objectively, they’re more designed. Apple makes more of the choices than do PC makers."
Is this a very western-centric view? Maybe... But it seems like everyone wants things to just work and unless they are VERY price-conscious, they don't want to think about it much.
The PC was successful because everyone wanted a computer that was like a PC, and two things happened: With the clone market, enough manufacturers could make the same thing, thus making computers affordable. Secondly, MS-DOS (and later, Windows) became a de facto standard preventing manufacturers from fragmenting the market with vendor-specific platforms. You bought a PC and it could run any DOS application.
Now, the clone market happened because the PC had standard components, which allowed any vendor to compete. Why can't standard components drive a new revolution in phones? Because everyone already has cheap, disposable phones. As far as the consumer is concerned, the mobile phone's "PC moment" already happened, except without the modularity.
If there's anything to disrupt today, it's the network.
(Would a modular phone be nice? Sure, probably. Would it work, physically? Not if you want to keep the form factor of today's slim smartphones.)
This allows them to quickly assemble low-cost phones with components from a large number of manufacturers, driving down costs.
So the financial benefit that came from the open PC ecosystem has already been realised on cheap Android phones, the equivalent of the crappy PCs assembled from lowest bidder components of yesterday (Dell, Packard Bell, etc).
Think about laptops, they've been around for a lot longer and are much closer to desktop PCs. Why hasn't modularization transformed that market? What about tables?
Answer: Most consumers don't care. They won't upgrade their PC, their laptop, their phones, their tablets, they'll just check them out and get a new one every 5 years.
Phones and tablets don't have enterprise demand. Laptops have some enterprise demand, which corresponds to some degree of customization/standardization.
A phone has nearly zero empty space. It has to be very compact, especially thin, it has to have a large battery, a large screen, and powerful chips. Also consider things like antennas, microphones, speakers, usb and headphone connectors that cannot be made smaller than physics and standards dictate.
A phone has no room for connectors between major components that would be remotely convenient for the user. Many connectors, is placed in accessible ways, would also work against mechanical strength. Good connectors that won't break after 2-3 reconfigurations, aren't cheap. Thus a phone with connectors will need to be significantly larger and heavier, and also costlier.
The PC argument does not work with mobile devices. At best, major components, like radio interfaces board, or a screen, or a battery can be made standardized and user-replaceable, likely requiring tools and skill. Batteries are easiest, but even these are far from standardized, and there are little incentive for manufacturers to work towards this. Users are content with glued-in batteries anyway.
But one or two expansion slots... now that would've been revolutionary. You could add whatever mattered to you: medical devices, specialized cameras, etc. etc. etc. Instead we're left with shitty USB or audio jack dongles or detached Bluetooth devices. Suck.
Everything that would be used by more than 5% of users should still be integrated, everything else would be a waste of resources because high numbers have such a strong effect of making electronics cheap and light.
How is USB not a fantastic answer to the problem you described?
USB is a decent electrical interconnect and API standard... But not a good cellphone mechanical modularity standard.
Also, MicroUSB on a lot of devices doesn't seem like the most robust port ever - I think it would eventually wear out with more than just the nightly plug/unplug for charging.
Many, especially in the 90's when compelling software often necessitated hardware upgrades.
Besides, public opinion on computing as a platform seems all but dead. This is a computers-as-appliance age, where everyone is punch-drunk on services and single-use apps. Power users seem to be disappearing; nobody wants to learn how to do anything and so much of what sells are just variations on a it-wipes-my-own-ass theme. Platforms don't work in that world; too few want to bother with them.
Nobody wants to simply build computers anymore, it seems. Too boring.
(edit: On the margins thing-- the big three enjoy greater margins because they offer greater services, insulating them from said pressure)
Allowing interoperability on your system allows other vendors to drive down your margins. That's why we're in the mess we're in. It's part of the problem with IoT, too. It would be great to have a secure, interoperable set of home automation parts, but every manufacturer wants you locked into their cloud and buying new devices every few years to maintain the revenue stream.
Ever increasingly cheap and plentiful clones made computers finally cheap enough for anyone to own?
Phones are already way beyond that point.
> where if a user needs more ram he would just add more, if he needs a new processor he would just replace the current, if he needs a higher resolution display the old just would get replaced.
Why would I ever want to deal with that?
No one wants to deal with that. Just works is way better than endless tweaking.
Because it's cheaper and easier than throwing out a phone and getting a new one. Phones could less 10 years with ad hoc upgrades instead of 2. I'd love to have been able to take my galaxy nexus (best phone ever) and add more ram and a faster processor.
Being made largely out of off the shelf parts, and coming from a company with a massive brand recognition in the business world, the PC was ripe for cloning.
The standardization only happened after the BIOS had been cloned, and the clones defended successfully in court.
As Project Ara moved forward, it became more and more clear that Google would have none of that clone business.
How has google not already beaten the iPhone? They have vastly more users, and have very small margins (the end state for PCs)
By market share, Google won years ago. By profit, commoditizing the phone hardware even further would drive android manufacturer profit down even further.
I even think this is how Google would have beaten the iPhone. Android as the OS for an ever-evolving phone, where if a user needs more ram he would just add more, if he needs a new processor he would just replace the current, if he needs a higher resolution display the old just would get replaced. Of course that is not easy, of course in the beginning a tailor-made smartphone would be slicker, faster. But then you support it some time and you get loyal users upgrading their device instead of even considering something else.
I know there are limits to upgradeability, just as a PC reaches the moment where all that remains from its old incarnation is the case. But still, enabling upgrades and thus removing consumers from the market would be a huge hit to all other smartphone vendors, probably spawning a more targeted parts industry.
And no: The current module system from LG or what Ara was presenting in the end is not it. Adding a new photo lense interests no one. You need to be able to switch out the core components to make an impact. The Fairphone 2 comes close, but I don't think you can drive real core upgrades without controlling the OS, and thus it ends up with a core module. We'll see. If they manage to add things like replacing the display with a better display, the core module with a more powerful one, that might already be good enough.