Okay, this isn't a Linux question and it's actually pretty dumb to ask. I know this isn't possible, but can anyone explain why it's not possible to split an ethernet cable at one end for two machines to share one port on a switch or hub. I know that it's easier to just run two cable, but is this not possible because TCP/IP will get confused?

Sorry for the silly question. It's a little debate at work and before we get stupid and decide to make cable I just want to know why it's not going to work... lol

I am not sure I understand your questions, but when you say "split" the cable, I can give you a general response to why it is not ok to split a cable. (unless connecting it to a patch panel or something)

It's twisted pair, meaning that all the wires have to be twisted in order to reduce electromagnetic induction until the very end connection, else the packets running through the cables will be all messed up.

You can't split the cable because the two computers will be trying to talk on top of each other. With ethernet, two pins are used for transmit and two for receive. Hubs and switches reverse their own pins (which is why directly connecting two computers requires a crossover cable) to connect to the computer.

Ethernet is based on the principle that only one device feeds into a cable, put simply.

A lot of companies are using this and it works fine:
http://www.cabling-design.com/refere...ts/split.shtml

It only works for ethernet that uses 2 pairs, I can't remember cabling of Gigabit,.. I'm sure others know for which ethernet it doesn't work. Oh yes, you can't put a DCE (hub,switch,router,..) between the hack!

Bit Off topic, this crossover cabling image
http://www.itkomi.ru/gallery//displa...bum=18&pos=235

Actually, that's backwards. Ethernet is a shared medium by design. In essence, all a hub does is to tie all the wires together and let all the machines talk on the same exact set of wires! The machines themselves are responsible for re-transmits and attempting to avoid overrunning each other (IIRC, the CDMA/CS protocol is used).

A switch on the otherhand changes things...

You could push the boat out a bit and grab some CAT 8, some EC7's and hey presto, two ethernet devices on one cable.

Yes, but the OP only wanted to use one port on the hub/switch, which that would not work for.

WHile RJ45 has 8 pins and Cat5 structured cabling wires all the pins through the actual Ethernet interface in most cases only uses two pairs, pins 1,2,3 and 6.
There were a few early flavours of 100M ethernet that used all 4 pairs but they disappeared quite quickly.

This is only true of Co-axial versions of ethernet. Twisted pair ethernet is not electrically designed as a bus like the co-ax versions. So while the CSMA-CD access method means that all ports on a repeater are in fact a single ethernet segment, this is not electrically true of the twisted pair transmission medium, only due to the behaviour of the hub (which is indeed simply a multi port repeater).

It is perfectly possible to run 2 pcs to two ports on 1 cat5 run by using 2 pairs for each, but just make sure that you don't split the pairs as this will cause you all sorts of problems.

By and large its best avoided though.

That is also going to create all kinds of noise and crosstalk on the cable(probably).

soule

Why more so than two cat 5 cables that are laid next to each other? All that separates the pairs is a milimeter of pvc. UTP is unsheilded after all.

I run two uplinks over 50 meters down to my garage on one cable this way, runs clean as a whistle.

The whole point about twisted pair transmission is that it uses differential transmitters and receivers. The pair is twisted so that both wires occupy the same position in space as near as possible. Any noise affecting once wire must affect both inn the pair equally. The differential receiver filter out any signals on the pair that are common. So no cross talk is not an issue in this case.

The main problem is splitting the pairs. This WILL cause you problems.

You could connect 2 PCs to the same port but communications would be limited to half duplex since the adapter would need to sense collisions (CSMA/CD, or Carrier Sense Multiple Access/Collision Detection). You are also probably limited to 10MB.

I haven't found anything specific if you can connect two or more receivers to the same transmitter. I've found several schematics for passive hubs that use diodes or ICs to isolate the PCs network adapters. 10base2 used coax that with all adapters connected together.

Look , you cannot electricaly connect two transmitters to the same receiver in UTP. Its not electrically a bus. The transmitters do not use tri-state outputs and therefore cannot float.

Also the transmitters aren't passive in an idle state - a special IDLE code-group is continuously sent during interframe gaps to maintain continuous synchronization between the NICs at each end of the link. The receipt of IDLE is interpreted to mean that the link is quiet.

If one transmitter asserts a signal with +2.2v across the pair and another asserts -2.2v across the pair (neither of these can be called one or zero due to the differential manchester encoding) you get 0v differential total. I.E you just receive crap. Its simple basic electronics.

Neither NIC would ever succesfully pass Fast Link Pulses, let alone actually transmit any valid data.