Intel's Light Peak is not the only connection answer

Sometimes less choice is better. Throughout the history of computing, there have been almost as many hardware interfaces as there have been manufacturers. A socket and plug for each different type of interface. And it’s time for this to end. It’s time for an interface that, for better or worse, can do everything.

At the time of writing, the standard interfaces include, USB, FireWire 400, FireWire 800, Display Port, DVI, HDMI, RJ45 Ethernet, SATA and eSATA. Those are just the common ones. Now Intel is getting ready to deliver Light Peak so there’s going to be yet another hardware interface. Another different set of leads and plugs.

From a consumer perspective, the spaghetti just gets worse. The chances of having the right lead for a particular job is dismal. USB made some good headway but it’s just not flexible enough to do everything. USB 3 is, in theory, just about fast enough to deliver a 1080P60 signal to a monitor. However, it has awful cable length limits.

FireWire beats USB in almost every respect. Except that it’s as good as dead. SATA is too flimsy and eSATA far too narrow a specification. DVI has too many pins. Display Port and HDMI are where it starts to get interesting.

You can send an Ethernet signal over the latest HMDI 1.4 connections. An HDMI lead can be extended with a standard bit of CAT5e or CAT6 network cable and an appropriate adaptor. This seems to have gotten some people interested enough to the point where there’s a new proposal that mixes the best of RJ45 Ethernet and HDMI called HDBaseT.

In some ways it might seem like the perfect answer. Make everything HDBaseT and it could all talk together. Everything would have high bandwidth if it needed it. Everything would have networking. Surely this is the best possible answer? Replace all of those old ports with HDBaseT.

Except that there’s one big problem: HDBaseT is rooted in the past. While you can send 10Gbps down it, that’s limited to the video signal. Networking on it only does 100Mbps. The fact that there’s no need to send anything except a 10Gbps network signal doesn’t seem to have occurred to them. With a little tweak, the chips that are capable of decoding an HDCP signal could easily decode packets of TCP/IP over Ethernet to get that signal.

HDBaseT is also rooted in the past through its choice of connector. The RJ45 has certainly stood the test of time but it has also been surpassed in almost every way. Plugs that lock in place may be great in a patch panel but they’re not designed to be plugged and unplugged regularly.

That leaves a simple and obvious conclusion. If you want to replace every lead out there, the answer is to give Ethernet a new connector. It solves almost every cable problem. It also gives a whole new list of possibilities.

Going back to the earlier list of standard connectors, 10GBaseT Ethernet with PoE is capable of replacing every single one of them. USB and FireWire don’t have the power carrying capabilities or cable length. Display Port, DVI and HDMI can’t do networking anything like as well. SATA and eSATA have less bandwidth and can’t carry power.

There are the obvious arguments against. Ethernet switches are more expensive than USB hubs. Chips capable of 10GBaseT have high clock speeds and are also expensive. But economies of scale would drive the prices down hard and fast.

The only thing that comes close is Light Peak. That’s capable of 10Gbps and is meant to do exactly what’s described here. It’s meant to replace every lead. It’s also meant to scale to 100Gbps eventually. But it probably can’t carry power in the same way. And all of the arguments that go against using Ethernet everywhere also go against Light Peak. A chip that costs a lot to send a 10Gbps signal down optical fibre is going to be just as expensive as one that sends a 10Gbps signal down copper wires. Except that Light Peak isn’t a standard yet so it’ll probably be more expensive initially.

And there is where 10GBaseT wins. Ethernet switches are capable of handling different speeds going into them so you could put a slower 1GBaseT chip in a device that needs to be cheap. Or even a 100BaseT chip for very, very cheap. You’d probably have to plug your monitor directly into a machine if you didn’t want to hose your network but you could have the option to plug it in somewhere else.

Imagine the effect on the average consumer of being able to plug in any peripheral into any port because all of those ports are the same. Even inside the machine. Ethernet has been with us for years so maybe the technology to do it has been with us for all of that time. Maybe it’s just had the wrong plug on it. Or maybe Light Peak will be what kills copper wires.