When building a small home ethernet network, we normally use switch to connect all cables together. I remember 10 years ago it was also possible to buy a much cheaper device called hub which can be used instead of a switch on 10 mbit/s UTP based network. When I read hub article in Wikipedia, they say it is an active device. My question is: Why is a hub an active device, or is each hub really active? Hubs I used were simple devices without power supply and they simply connected all cables together (with a few diodes or something logically equivalent). So in my eyes it is a passive device, or I don't clearly understand the meaning of active/passive here.
Ethernet hubs can be passive, however they are limited in connectivity range by the signal loss through cabling and the hub. There are designs available to build them yourself. I don't know of any current manufacturers that make them since switches have made hubs obsolete.
Hubs were always active as they contain powered electronics. I've never seen one that was completely passive requiring no extra power source to make it function.
The difference between a basic switch and a hub is that a hub broadcasts the packets to all other links. A switch is smarter and only sends packets to links that require it. So switches will always need to be powered to run their logic chips.
But why do hubs need to be active? One of those reasons that comes to mind is to ensure that they meet the performance requirements of the 10baseT standard which is 100m. If it was entirely passive like 10base2 you run into distance limitations as the the wire resistance and capacitance increases with length and loading. You can't meet these requirements with a purely passive design. Therefore each Ethernet link would need it's own active driver.
There used to be old style ethernet which was 10base2 which acted passively. .
The disadvantage of this being that if you break the cable you break the connection for potentially all computers. It was as I recall somewhat unreliable compared to what we have now.
That being said, 10Base2 had a 200m length limit and up to 30 nodes. But look at the wires, they are very much thicker then 10BaseT. Also you can't buy 10Base2 anymore and it was limited to 10mbit.
Modern copper ethernet uses 10baseT (t for twisted pair) for it's cabling and is much more reliable.
They are active because its not like connecting everybody's signals together. They have a processor, memory, they receive, enqueue and forward packets according if its possible to transmit them (avoid colisions). Switches will also memorize which MACs are present on each port so they know which port is to send data to that MAC. They take decisions.
This behaviour also make them sort of work as amplifiers. A CAT-5 cable will work for a maximum distance of 100 meters (328 feet), but if you put a HUB or Switch after 100m you can add another 100m and keep going farther whatever distance you want. I know its not a exclusive characteristic of active devices, but its worth mentioning.
Howover, the traditional Switch is not an "active device on the network", they don't have an address, they cant be "accessed" or "pinged", they are silent traffic wardens managing the network in the data link layer, but being silent doesn't mean they are passive at all.