8 ms·

It cannot be emphasized enough how much the recent USB specifications are dropping the ball. I wouldn't mind paying a small premium just for a guarantee that th

It cannot be emphasized enough how much the recent USB specifications are dropping the ball. I wouldn't mind paying a small premium just for a guarantee that things will Just Work when I connect them. Instead, USB device manufacturers want to cost optimize everything, which probably forced the standard to become so Byzantine. Understandable from a manufacturer's point of view, but terrible from an end user perspective.

dwaite 6y ago
I blame a combination of poor enforcement of trademarks and a desire to let people support the latest specifications with the cheapest possible build cost. The poor enforcement of trademarks being - that there should not be a plethora of things claiming to be USB which are not spec compliant, nor should people be using names like "USB 3.2 Gen 1x1" vs the marketing name "Superspeed USB". You also see this poor enforcement with other organizations, such as DisplayPort vs DisplayLink. This spills over to problems with cabling, such as how much wattage is supported for power delivery or what speed data transfer is available. Oddly Intel is the one stepping up here with Thunderbolt 4, because it is now more of a certification than a specification.
- alacombe 6y ago
  You can't blame the lack of "enforcement" to be the cause of the ills created by bad standards. I'm fairly certain that 2 official certified USB-3 cables can brick devices even though the physical form factor is the same. Relying on cables internals for function is just as stupid as can be.
  - eigenvector 6y ago
    If you think USB is bad, try looking at all the various things that can be connected together with M.2 slots vs. the pairs of devices that will actually work together despite having the same slot. I have a degree in electrical engineering and after years I only just realized that there are actually PCIe M.2 storage devices that only support AHCI mode not NVMe.
    - wtallis 6y ago
      > I have a degree in electrical engineering and after years I only just realized that there are actually PCIe M.2 storage devices that only support AHCI mode not NVMe. To be fair, there were only a handful of those really early on when M.2 was just getting started and OS/BIOS support for NVMe wasn't universal yet. And hardly any of them were released as retail products; they were mostly OEM-only drives. All of those drives went out of production 3-4 years before the first host devices that require NVMe and can't work with AHCI started to show up: USB to NVMe bridge chips for external enclosures. So if you haven't encountered an AHCI M.2 SSD yet, you probably never will and knowing about them is just an obscure historical curiosity.
  - andromeduck 6y ago
    I think USB-C would have been so much better if the CC system were saner - for example we could have had a shift uart like register based api: - Host queries some property and each cable along the way appends their value to that message and passives it along. - The endpoints either loopback if they're the new standard or pull down the pins if they're USB-2 only. - It follows that if no message makes it back then we're limited to USB 2.0 and if a message does make it back, we can query/configure all elements along the path of connections. Right now this is impossible which makes it impossible to detect if there is a non 50v tolerant hubs between two hubs. - The programming/hw implementation for PD and alternate mode would also be non-insane in this model. Currently PD requires stupidly complex state machines, op amps, resistor banks or some autonomous IC to pull off and there are so many screwups because it's such an obtuse standard. - Supporting alternate modes, and providing more diagnostic information would be much easier. There should have only been 2 passive cable types - 2.0+PD and USB 3 gen 1 - everything else should require a smart tag on the chip.
    - Pet_Ant 6y ago
      > CC system What does "CC" stand for?
      - InitialLastName 6y ago
        Configuration Channel There are an added pair of pins/single wire in USB-C cables that allow the devices to a) detect the orientation of the plug and b) do extended power negotiation for the increased power capabilities of USB-C.
    - ethbr0 6y ago
      This so much. The current USB @&$#-up seems to be the result of "EEs try to expand line-level encoding to create an API." This should all be solved at the negotiation layer, even if that needs to be made more complex, so that the remaining components can be simpler and reasonably-behaved. Instead, we got something that allows each device to be a bit more electrically simple, at the cost of ballooning complexity for the ultimate use case. USB-IF took their eye off the ball, and wrote a spec for manufacturers, without thinking about the consequences to consumers. At some point, it's a value trade-off between {working product for use cases} and {+$2 on BoM}.
etrabroline 6y ago
>USB device manufacturers want to cost optimize everything, which probably forced the standard to become so Byzantine As Intel loses its raw CPU performance advantage, it's new marketing materials promote "features" and "convenience". Practically speaking, USB is defined by Intel, and it is making a visible choice to allow itself a position in the market for Thunderbolt 4 by making some USB4 features optional so they can include them in TB 4, and adding other small minimum performance guarantees and hoping that is enough to sell TB 4 chips to high-end AMD motherboard makers, which in turn makes their own CPUs (with built-in thunderbolt 4) more price competitive. It is a sound business plan, but not in the best interest of consumers.
- selfhoster11 6y ago
  > It is a sound business plan, but not in the best interest of consumers. The fate of a universal lingua franca connector is too important to be left to just one company. Why is it still being defined by Intel rather than a panel?
  - dijit 6y ago
    It isn’t just Intel, there is the non-profit “USB implementers forum” (or, USB-IF) which oversees the specification. Intel, Microsoft, Apple and even HP are notable members. https://en.wikipedia.org/wiki/USB_Implementers_Forum https://en.wikipedia.org/wiki/USB_Implementers_Forum
alacombe 6y ago
> It cannot be emphasized enough how much the recent USB specifications are dropping the ball. This outcome was pretty obvious, it's literally in the name of the damn thing: UNIVERSAL Serial Bus. The only way to achive that is to lower yourself to the lower common denominator.
- royjacobs 6y ago
  But that's the point, right? For a manufacturer "universal" means the lowest common denominator whereas for an end user "universal" means...well, universal.
  - alacombe 6y ago
    ... as long as there is no confusion possible for the end-user.
DrBazza 6y ago
I lived through the pre-USB world, and things were even more "fun". In all seriousness though, the current state of a USB is ridiculous. My wife innocently asked me why her USB C phone isn't triggering Android Auto in the car, and straight away you know it's because she bought a USB C power-only cable without knowing the difference. Nice.
- bzb6 6y ago
  Aren’t power-only cables a security feature?
  - zdkl 6y ago
    They are, and frankly that they need to exist is the single most egregious example of why USB isn't what I need it to be. What (I'm surprised) I haven't seen is a male-to-female adapter with a switch on it that physically kills the data tx lines.
    - AussieWog93 6y ago
      A device very similar to the one you described exists: the aptly named "USB Condom"!
      - lasereyes136 6y ago
        Indeed, I have two of them. They don't always work correctly. Meaning they don't always charge depending on the device and cable you plug into them.
        SAI_Peregrinus 6y ago
        I've got a (USB2) power monitor that includes a chip (just an STM8L051F3 microcontroller) to manage power negotiation and keep data disconnected. So it negotiates power with host and device separately, without ever needing to connect the data lines from host to device directly. I expect there are or will be versions for USB 3.<whatever I've lost track>.
qppo 6y ago
Consumers want one cable and connector for all devices regardless of the ridiculous notion that all devices should have similar enough electrical requirements that they could use one cable. The result is USB.
gumby 6y ago
It's no accident that the USB group is called the "USBIF" -- "USB Implementor's Forum". This is different from, say, the C++ standards committee which is a combo of compiler & std library developers (i.e. implementors!), enthusiasts, and some who are best referred to as "users". Not a criticism of the USBIF* just pointing out the motivations to underline your point. * BTW I have plenty of criticism, this just happens not to be an example.

dwaite 6y ago

I blame a combination of poor enforcement of trademarks and a desire to let people support the latest specifications with the cheapest possible build cost. The poor enforcement of trademarks being - that there should not be a plethora of things claiming to be USB which are not spec compliant, nor should people be using names like "USB 3.2 Gen 1x1" vs the marketing name "Superspeed USB". You also see this poor enforcement with other organizations, such as DisplayPort vs DisplayLink. This spills over to problems with cabling, such as how much wattage is supported for power delivery or what speed data transfer is available. Oddly Intel is the one stepping up here with Thunderbolt 4, because it is now more of a certification than a specification.

alacombe 6y ago
You can't blame the lack of "enforcement" to be the cause of the ills created by bad standards. I'm fairly certain that 2 official certified USB-3 cables can brick devices even though the physical form factor is the same. Relying on cables internals for function is just as stupid as can be.
- eigenvector 6y ago
  If you think USB is bad, try looking at all the various things that can be connected together with M.2 slots vs. the pairs of devices that will actually work together despite having the same slot. I have a degree in electrical engineering and after years I only just realized that there are actually PCIe M.2 storage devices that only support AHCI mode not NVMe.
  - wtallis 6y ago
    > I have a degree in electrical engineering and after years I only just realized that there are actually PCIe M.2 storage devices that only support AHCI mode not NVMe. To be fair, there were only a handful of those really early on when M.2 was just getting started and OS/BIOS support for NVMe wasn't universal yet. And hardly any of them were released as retail products; they were mostly OEM-only drives. All of those drives went out of production 3-4 years before the first host devices that require NVMe and can't work with AHCI started to show up: USB to NVMe bridge chips for external enclosures. So if you haven't encountered an AHCI M.2 SSD yet, you probably never will and knowing about them is just an obscure historical curiosity.
- andromeduck 6y ago
  I think USB-C would have been so much better if the CC system were saner - for example we could have had a shift uart like register based api: - Host queries some property and each cable along the way appends their value to that message and passives it along. - The endpoints either loopback if they're the new standard or pull down the pins if they're USB-2 only. - It follows that if no message makes it back then we're limited to USB 2.0 and if a message does make it back, we can query/configure all elements along the path of connections. Right now this is impossible which makes it impossible to detect if there is a non 50v tolerant hubs between two hubs. - The programming/hw implementation for PD and alternate mode would also be non-insane in this model. Currently PD requires stupidly complex state machines, op amps, resistor banks or some autonomous IC to pull off and there are so many screwups because it's such an obtuse standard. - Supporting alternate modes, and providing more diagnostic information would be much easier. There should have only been 2 passive cable types - 2.0+PD and USB 3 gen 1 - everything else should require a smart tag on the chip.
  - Pet_Ant 6y ago
    > CC system What does "CC" stand for?
    - InitialLastName 6y ago
      Configuration Channel There are an added pair of pins/single wire in USB-C cables that allow the devices to a) detect the orientation of the plug and b) do extended power negotiation for the increased power capabilities of USB-C.
  - ethbr0 6y ago
    This so much. The current USB @&$#-up seems to be the result of "EEs try to expand line-level encoding to create an API." This should all be solved at the negotiation layer, even if that needs to be made more complex, so that the remaining components can be simpler and reasonably-behaved. Instead, we got something that allows each device to be a bit more electrically simple, at the cost of ballooning complexity for the ultimate use case. USB-IF took their eye off the ball, and wrote a spec for manufacturers, without thinking about the consequences to consumers. At some point, it's a value trade-off between {working product for use cases} and {+$2 on BoM}.

etrabroline 6y ago

>USB device manufacturers want to cost optimize everything, which probably forced the standard to become so Byzantine As Intel loses its raw CPU performance advantage, it's new marketing materials promote "features" and "convenience". Practically speaking, USB is defined by Intel, and it is making a visible choice to allow itself a position in the market for Thunderbolt 4 by making some USB4 features optional so they can include them in TB 4, and adding other small minimum performance guarantees and hoping that is enough to sell TB 4 chips to high-end AMD motherboard makers, which in turn makes their own CPUs (with built-in thunderbolt 4) more price competitive. It is a sound business plan, but not in the best interest of consumers.

selfhoster11 6y ago
> It is a sound business plan, but not in the best interest of consumers. The fate of a universal lingua franca connector is too important to be left to just one company. Why is it still being defined by Intel rather than a panel?
- dijit 6y ago
  It isn’t just Intel, there is the non-profit “USB implementers forum” (or, USB-IF) which oversees the specification. Intel, Microsoft, Apple and even HP are notable members. https://en.wikipedia.org/wiki/USB_Implementers_Forum https://en.wikipedia.org/wiki/USB_Implementers_Forum

alacombe 6y ago

> It cannot be emphasized enough how much the recent USB specifications are dropping the ball. This outcome was pretty obvious, it's literally in the name of the damn thing: UNIVERSAL Serial Bus. The only way to achive that is to lower yourself to the lower common denominator.

royjacobs 6y ago
But that's the point, right? For a manufacturer "universal" means the lowest common denominator whereas for an end user "universal" means...well, universal.
- alacombe 6y ago
  ... as long as there is no confusion possible for the end-user.

DrBazza 6y ago

I lived through the pre-USB world, and things were even more "fun". In all seriousness though, the current state of a USB is ridiculous. My wife innocently asked me why her USB C phone isn't triggering Android Auto in the car, and straight away you know it's because she bought a USB C power-only cable without knowing the difference. Nice.

bzb6 6y ago
Aren’t power-only cables a security feature?
- zdkl 6y ago
  They are, and frankly that they need to exist is the single most egregious example of why USB isn't what I need it to be. What (I'm surprised) I haven't seen is a male-to-female adapter with a switch on it that physically kills the data tx lines.
  - AussieWog93 6y ago
    A device very similar to the one you described exists: the aptly named "USB Condom"!
    - lasereyes136 6y ago
      Indeed, I have two of them. They don't always work correctly. Meaning they don't always charge depending on the device and cable you plug into them.
      - SAI_Peregrinus 6y ago
        I've got a (USB2) power monitor that includes a chip (just an STM8L051F3 microcontroller) to manage power negotiation and keep data disconnected. So it negotiates power with host and device separately, without ever needing to connect the data lines from host to device directly. I expect there are or will be versions for USB 3.<whatever I've lost track>.

qppo 6y ago

Consumers want one cable and connector for all devices regardless of the ridiculous notion that all devices should have similar enough electrical requirements that they could use one cable. The result is USB.

gumby 6y ago

It's no accident that the USB group is called the "USBIF" -- "USB Implementor's Forum". This is different from, say, the C++ standards committee which is a combo of compiler & std library developers (i.e. implementors!), enthusiasts, and some who are best referred to as "users". Not a criticism of the USBIF* just pointing out the motivations to underline your point. * BTW I have plenty of criticism, this just happens not to be an example.