Half-duplex connections were popular at the time of 10Mbps and 100Mbps Ethernet, and, according to standards, it is allowed also in case of 1Gbps Ethernet.
Am I correct that half-duplex mode support in Ethernet chipsets was crucial in case either a network hub (hub is internally a single wire) or some other shared Ethernet medium (for example, 10BASE-5) was used?
Are there any reasons for half-duplex connections in Ethernet environments where twisted-pair cabling is used and hubs are not used?
The reasons for half-duplex ethernet are as you understand them. In fact, there was a movement to not include half-duplex for 1000Base-T, but it still made it into the standard. For 10 Gb ethernet, half-duplex was dropped so there is no such thing as 10 Gbps half-duplex ethernet as a standard.
Unless you still have a hub (they are still around) or a device that doesn't support full-duplex (they exist, especially for 10Base-T), 10Base-T or 100Base-TX on UTP don't really need half-duplex.
Ethernet was derived from https://en.wikipedia.org/wiki/ALOHAnet which used UHF radio; an inherently half-duplex medium when operating on the same frequency.
Early wired networks used a coaxial bus until well into the 1980s.
In the case of partial wiring damage, half duplex can be the difference between a slower connection and NO connection (until the damage is repaired.)
That can be a BIG difference.
Rich Seifert gives the answer (for Gigabit Ethernet) in his book "Gigabit Ethernet" and in an usenet post:
Quote:
The answer is more political than technical. Gigabit Ethernet was developed under the auspices of the IEEE 802.3 Working Group. By definition, 802.3 networks must include the capability of CSMA/CD operation. [Note: This was true at the time of the writing of the Gigabit Ethernet stamdard; it is no longer the case.] If Gigabit Ethernet offered a full-duplex-only solution, it would have been difficult to justify its development within the IEEE 802.3 Working Group. [...]
- The resulting standard would have had difficulty calling itself “Ethernet,” since it would not use CSMA/CD (even as an option) and it would not have been developed as part of IEEE 802.3, the recognized “owner” of the Ethernet name.
The reason for half duplex, was due to cost, initial 10Meg ethernet shared the same coax cable only two conductors 10Base2. At the time it was not imagined that any thing as fast as 10Mbit/s could be transmitted on a unshielded cable.
One port on a switch/hub could have 20 pc's connected in daisy chain, thus CDMA was used to listen for traffic, start sending, back-off for random time if someone started transmitting at the same instance and send again.
This is also the reason for the length restriction and minimum packet size, as you had to be sure before the last bit was sent that someone on the other end of the same cable did not also start to transmit at the same instant.
As electronics became faster and cheaper, we moved to Cat5, individual cables/ports for each pc, and then to switches, where the packet can be stored and forwarded in both directions at the same time (Full Duplex)
Am I correct that half-duplex mode support in Ethernet chipsets was crucial in case either a network hub
Yes
hub is internally a single wire
It would be more accurate to say a hub simulates a single wire, it's actually a bit more complex than that.
or some other shared Ethernet medium (for example, 10BASE-5) was used?
Yes
Are there any reasons for half-duplex connections in Ethernet environments where twisted-pair cabling is used and hubs are not used?
The other big one is old end devices.
10 Megabit gear is unlikely to support autonegotiation, so full duplex mode cannot be selected automatically. You could potentialy force both the device and the switchport to full duplex mode but doing that risks duplex mismatches happening down the road, so it's usually more trouble than it's worth.
