I say "in theory" because, as with any other technology, Ethernet has its limits, both physical and software — or, more accurately, higher-level protocol-based. The protocols aren't strictly Ethernet, which comprise the bottom layers of the OSI model; but, without the protocol to interpret the message carried by the Ethernet layers, it would all be noise.
The most obvious physical-layer limitation is that Ethernet is generally limited to a maximum spur length of 100 m. In many cases, this will not be a factor but to connect a device more than this distance away from your interface room, it will be necessary to install additional hardware to clean up and boost and signal back up to acceptable levels before retransmitting it another 100 m. Also, the maximum number of repeaters is limited to four, unless you switch to optical fiber. But fiber is not PoE-capable, and, of course, you then need a reliable power supply at each of the repeater stations. Otherwise we will have the perpetual-motion-machine conundrum — energy for nothing. Fortunately, with the trend toward distributed interface rooms, this average difference continues to decrease.
Industrial-grade PoE is gaining traction with products available from many manufacturers. And, because they are compliant with 802.3af, they're interchangeable and able to connect with each other or components from the business environment, just like other Ethernet products.
The range of PoE network equipment also includes splitters — separate power and signal for those devices unable to detect PoE or to tap into the power — and injectors, which add power to a network from somewhere other than the end device. The increased interest in wireless networks and need for associated distributed wireless access points seems to be the logical application that will drive the adoption of industrial PoE.
A second challenge to PoE for automation applications is that the 802.3af standard specified a voltage of 48 Vdc, while this industry standardized on 24 Vdc. At least one industrial Ethernet equipment supplier has a PoE splitter with a 24 Vdc output.
Another physical-layer limitation, in about 20% of installations, is transmitting Ethernet into a classified area. There always is the explosion-proof enclosure option, but it also is possible to go with an intrinsically safe solution from MTL and Controlled Systems providing up to 500 mA power over an IS RJ45 cable. This solution isn't an open standard, although there is some discussion within the British Standards Institute (BSI) about trying to push something through the IEC in this area.
Hardware and the associated physical signal are only one part of the communication equation; the other is the protocol or language used. Unfortunately, a single standard is unlikely. Just as with fieldbus protocols, it is likely we'll have protocols divided by industry, making it tougher for end-device and control-system manufacturers to justify the necessary development expense.
We are close. Hopefully Ethernet field devices will soon enter the market in quantity, and not just to do the same signal at higher bandwidth, but as an enabler to new and innovative ways of doing things or making measurements. My favorite new application is a distributed spectrometer similar to a thermowell with an LED for illumination, diode array to capture the data and Ethernet to send it back to a central system for processing. No sample system or moving parts.
PoE has removed the single cable to a device problem, but minor limitations remain to be resolved. PoE is not yet the silver bullet or magic elixir, although, hopefully soon, it will get closer, but that's another story for another time.