Hacker News Reader

Comment by userbinator | Hacker News Reader

In an alternate world, Ethernet took on the role of the universal serial bus, and we have laptops that charge via PoE, but only possible on one of their ports (the others are usable for peripherals --- with protocols running over Ethernet too, of course.) But the same confusion regarding power and speed capabilities exists.

View on HN

We'd have to invent a new connector first. It's too thick for modern laptops, not to speak of cell phones.

Also, RJ45 is terribly fragile if you keep plugging and unplugging it, eventually that latch will break. And copper can barely support 10G and is terribly power hungry when it does that. And the cables get thick and inflexible.

The 8 pin modular connector as found in most ethernet does have several sins but it has one huge redeeming feature, A feature I wish was found in every cable. It is easy to field terminate. Have fun putting a new end on nearly any other cable.

Field termination is necessary when the connectors are too large to pull through a conduit. But if they were USB-C sized, you could just pull fully assembled cables.

It’s also about managing length and slack.

It also comes in very handy when you need a 8m cable, but only can buy them in lengths of 5m and 10m, or when you’re wiring an entire building, and figuring out which lengths to order up front is a major pain in the ass, certainly compared to ordering a few hundred meters of cable, a few hundred connectors and tools to put the two together. And that’s ignoring the price difference.

except those assembled cables will always be the wrong length.

As a person who has installed hundreds of miles of cabling of every description inside of buildings of every description:

Every single time someone has provided pre-terminated cabling for one of my jobs to "save time" or to "make it easier", this provision has done neither.

Instead, it has consistently multiplied both the time required and the installation difficulty. It has done these things while also producing an inferior end result.

It is my anecdotal observation that it's NFG.

I have no idea what kind of cable it was, but the bloke who installed the control panel for my ducted air conditioning got the cable snake stuck in the wall cavity. He had to cut it and use a different snake. So there's a dead snake in my wall, and your comment brought this to mind.

Lenovo has re-invented this particular wheel to fit in laptops, some ThinkPads come with a proprietary Ethernet port which is around the size of USB-C, just with Ethernet signals. And you can get a passive breakout adapter to convert it to RJ45 (idk if it's included with the laptop).

https://www.lenovo.com/us/en/p/accessories-and-software/cabl...

People routinely lose their cool over a $7 slim headphone to usbc adapter.

I wouldn’t mind dongles if their cable wasn’t so flimsy, especially when the other cables are strong.

Which they should, because that situation is a mess too. Either 1) you get an active dongle with a cheap and nasty DAC built in, or 2) you get a passive one and get to discover if your phone manufacturer decided to route the signal from the (probably quite nice) internal DAC out via the USB-C port or not; whether they did so competently given the proximity to signal and power lines; and - separately - whether it even has a mic in ring.

It really is fractally rubbish.

Just next to the proprietary port is a USB-C one. You can buy a good USB-C ethernet adaptor for less than half the price of the Lenovo dongle.

My ideal would be a non-proprietary, smaller native Ethernet connection capable of 10GbE.

The adapter still has to adapt. That requires power, adds cost, and adds negligible-but-non-zero latency. I don't love the proprietary port being proprietary, but the fact remains it is native Ethernet with no caveats.

They are getting rather small[0] and not-crazy-power-hungry these days!

And if we're being honest: is there really that much difference latency-wise between Ethernet -> NIC -> USB -> CPU and Ethernet -> NIC -> PCIe -> CPU?

[0]: https://www.jeffgeerling.com/blog/2026/new-10-gbe-usb-adapte...

The ix.industrial ethernet connector is a thing. I hate it, but it's a thing.

> copper can barely support 10G and is terribly power hungry when it does that.

AFAIK, thunderbolt cables are also copper - so what trickery do they use for supporting USB4-80? i believe both connectors use differential pair wires for signalling.

It's simply length. Ethernet is expected to work on 50-100m runs, while USB4 specifies maximum cable lengths of 2m even for just 5gbps (at least for passive cables). 80gbps is 0.8m

The longer Thunderbolt (which is actually just USB4) cables internally use fiber optics for data transmission, with converters to copper in each connector. Even the medium-distance (3 meter) ones have signal quality boosters in each connector matched to the kind of signal degradation that kind of cable will experience.

Completely passive TB4/TB5 cables max out at about 80 centimeters.

Besides what retatop and crote said - USB4 uses either binary or PAM-3, 10GBASE-T uses PAM-16. Higher modulation means lower frequency bandwidth (so longer and/or crappier cable), but also more current-hungry line driver and more current required to keep noise manageable.

> It's too thick for modern laptops

Nah, there's enough space for an RJ45 connector on the 0.48" thick E7270, so there's certainly enough space for one on the 0.61" Macbook Pro 14. The trick is putting the connector on the display hinge.

Laptops no longer come with ethernet ports because (a) wifi is good enough for most people, most of the time; (b) apple went USB-C-only in ~2018 and other 'premium laptops' copied it; and (c) by the time that trend reversed and laptops started re-adding hdmi and usb a ports, demand for ethernet connectors was lower than ever.

For what it's worth, it'd be pretty easy to design an RJ45 compatible connector that didn't have that flimsy plastic latch.

> copper can barely support 10G and is terribly power hungry when it does that

Mostly a side-effect of 10GBASE-T dating back to 2006 - and all the chips used by early 2020s prosumer devices therefore using pre-2010 technology.

Definitely a technological dead end, though - I highly doubt we'll ever see 25GBASE-T hit the market!

People would still complain that you can pickup the wrong cable and it won't work for 10GbE and that the ports look the same but some work on 10Mbit and others on 2.5GbE!

Some can even give and receive power and look the same as others that can't!

I modded a laptop to charge over PoE in 2007. Before realizing that the places that had PoE, and the places I wanted to charge my laptop, had nearly zero overlap. It was virtually useless in practice, but I still love the idea.

I have not yet made a laptop to output PoE. Though it would be tremendously useful for provisioning IP cameras, there are dedicated thick-tablet-shaped devices for that, which do source PoE from their batteries.

Even though both USB and Ethernet transport bits, the surrounding ecosystem is so different that it couldn't really be a replacement.

Devices plugged into an Ethernet network are true peers, but USB is master-slave by necessity. Ethernet devices have unique addresses, but USB devices can be anonymous, only identified based on the port they're plugged into. Ethernet is best-effort with buffering and packet dropping, but USB provides guaranteed delivery with tightly bounded latency. Ethernet signals must travel up to 100 meters but USB requires the host and device to be within a few meters. You could reuse the physical wires, maybe (we already do! USB runs on twisted-pair) but nothing else, from the connector to the topology, is usable.

I wish they would have developed a data-over-powerline protocol specifically to link your monitor's USB to your PCs. Especially for computers 20+ years ago, almost everyone plugged their monitor into the same AC power jack as the PC. They could have had a USB 1.x connection between the monitor -> AC power -> power supply -> motherboard so you could plug your slower USB devices into your monitor and skip the USB cable. Apple had this, I guess, with imac, and I know about powerline ethernet devices, but I think they skipped the most obvious use case.

That's because anything-over-powerline is absolutely and objectively terrible.

You already have a high throughput data cable between your PC and monitor. Carry USB over displayport or whatever. At least then you can use more than one PC on an entire city block.

I'd rather prefer Ethernet over USB 2.0. Full duplex, galvanically isolated and cheap fiber converters, (relatively) lightweight host implementation possible, no "transaction translator" madness (at the cost of slightly more expensive hubs). Too bad it is such a power hog.

Ethernet is different in part because it's for much longer cable runs (both average and maximum) than usb is. It's a lot easier to maintain signal integrity for 10 gbps or 40 gbps when you're dealing with a few meters maximum (0.8m max I guess for usb4?).

There are DAC cables for 10G and 40G (actually 4x10G) Ethernet which are only a few meters. The newer USB standards' physical layer and other characteristics actually resembles Ethernet in many ways.

Reply on news.ycombinator.com