I think I'm overusing commas...

My preference is, if there's a pause, use a comma.

+1 for commas...

I do find the statement that NetBIOS could only run over Ethernet to be somewhat dubious, though. I don't have a citation at the ready, but wasn't NetBIOS sometimes used over Token Ring? Also, might be worth clarifying that IPv4 can just as easily be carried over other PHY layers, like wireless Ethernet. (And then IPX was non-routable anyway...)

This separation between Ethernet and IP is actually useful because it gives flexibility – the same Ethernet can carry several different protocols, such as IPv4, IPv6, perhaps IPX - fun question: why can't we run TCP over Ethernet without IP in-between?

@MichaelKjörling - Yes... As strictly an API, NetBIOS is not a networking protocol. Older operating systems ran NetBIOS over IEEE 802.2 and IPX/SPX using the NetBIOS Frames (NBF) and NetBIOS over IPX/SPX (NBX) protocols, respectively. In modern networks, NetBIOS normally runs over TCP/IP via the NetBIOS over TCP/IP (NBT) protocol. This results in each computer in the network having both an IP address and a NetBIOS name corresponding to a (possibly different) host name. Source: Wikipedia

In 1985, IBM went forward with the token ring network scheme and a NetBIOS emulator was produced to allow NetBIOS-aware applications from the PC-Network era to work over this new design. Source: Wikipedia

@user20574 You might want to ask that as a separate question (fun questions are allowed!), but I'm pretty sure it's because TCP depends on a few things from IP, port numbers, ip addresses, checksum, etc. They're so closely intertwined they're often thought of as one thing, hence TCP/IP. You could make a different/more "bare" version of IP if you wanted and run it TCP -> BearIP -> Ethernet

@user20574: Sure, you could, but what's the point? You would miss out on all the features IP gives you. You'd have the same problems as I already described with pure Ethernet switching/bridging.

@shelvacu: Well, port numbers are part of TCP's own header, not IP's, so that's not a large issue. The protocols were designed at first to have clean separation of layers.

@MichaelKjörling: That was not my statement. Instead I was referring to the lack of routing features in NetBIOS. (My knowledge about NetBIOS is also rather minimal, but from what I know, NetBIOS-over-link-layer was much less useful than over-IPX or over-IP.) But as for IPX, well, it did have networks, routers, and even RIP; it was just non-hierarchical (so nothing like prefix summarization).

I remember about 12 years ago we had 768kbit DSL (digital subscriber line) and no routers. We had a 100Mbit hub and later a switch. We used to connect the DSL model to one port of the hub and have one computer dial in, which in Windows 2000 involved a PPPoE dial up connection, then share that with the regular network. That way, we could play with multiple computers online. We had to do that because there was only one networks card (yes, card, not onboard) so we could not hook up the modem directly. Good old times.

@simbabque: Ah, most ADSL modems here also had Ethernet-over-USB in case someone didn't have a NIC.

@grawity I don't think that existed back then. I'm talking Germany in 2002. My Deutsche Telekom DSL Model was big enough to need its own postal code. I had a 20m CAT-5 cable through the window to the basement where the NTBA was. That cable was on the modem, and that was hooked up to the hub. We used this to play Diablo 2 with two people on the internet at the same time. Worked sometimes.

@grawity "yes, static addresses are possible, but believe me, they're a pain in the ass to keep track of." I don't know about that. I implemented it on my home network and it makes it much easier and more reliable to deal with shared network resources such as NAS storage and printers. Once you set it, your are pretty much done. And if you have guest devices, they just get assigned DHCP without the hassle. The only thing you would need to administer for a guest is MAC filtering if you have that option set up.

@grawity I mention as this will likely be a post that many new users will refer to when they think about setting up a home network of their own, and they should be discouraged to undertake the little extra effort on the front end to set up static IP on their side of the network.

@AMR: Well if you have DHCP anyway, then what's the point of using static addresses for your own devices? Why not assing static leases instead?

@grawity It is possible that I used the terminology incorrectly and what I am doing is in fact DHCP Static Lease, where I assign a specific IP to each device on the router. an it inherits that IP address.

@AMR: Yes – an actual "static" configuration would be done directly on each computer by hand (i.e. without involving the router, as OP has described).