Carrier Ethernet... why utilize other transport mechanism?

Ethernet is not routable. Every switch on the network would need to maintain the list of all known etherhet addresses (MAC) so they would know which of its port should handle that packet.

Now, add in security issues, and you need to add VLAN tags with one VLAN per customer to make sure that broadcasts from one customer cannot be sniffed by another customer. (or prevent another customer from impersonating the MAC address of a device belonging to another customer and then getting/stealing its packets.

Another aspect is that telcos generally like to base new services upon existing infrastructure, so they create a LAN extension product that gets encapsulated into an existing protocol already supported by their infrastrcuture.

the 3 curses of the carrier.

1. inertia - many existing carriers have huge pervasive SDH networks already, and esp if Ethernet transport is "just another kind of circuit" it makes sense to combine it with all the other stuff.

SDH gets less and less efficient as the Ethernet circuits start to dominate the payload - the effect i see is that you get a mix of ethernet over sdh where there is spare capacity, but a parallel ethernet aggregation network grows initially in hot spots.

old joke - god would not have managed to create the world in 7 days if he had had an installed base......

1. scale. MPLS is proven in existing large backbones, and again in many cases it is already there.

1. common core network. L2 MPLS transport can use the same plumbing / core switches as a L3 VPN (and maybe an Internet backbone built using MPLS) - so 1 backbone can carry multiple services, which "should" reduce overall costs.

Also, MPLS seems to be 1 common choice for ethernet transport - probably because many carriers have already trained their engineers to look after it.....

already some good responses but i also like to point out that a basic ethernet network doesn't scale well either, I have had the unfortunate job of trying to stop a broadcast storm on a network with 300 odd layer 2 switches and inband management. By using a transport network to get your frames from A to B you can use a reliable transport method like SDH+DWDM to provide a basic layer 2 service to customers with all the SLA enhancements that you could offer with an SDH circuit.

Flamer.

Nothing is really converted, the Ethernet frame is just encapsulated into another frame type or just striped off and the underlying IP frame is encapsulated.

This is not an issue with the carriers because the pipe between A and B is just that, a pipe that moves bits from A to B. Even if the carrier had an Ethernet backbone network, the pipe would be virtual link created using either L2TP, or MPLS, or EoMPLS.

Again, not an issue, see above. In fact Cisco created private VLANs so that carriers could install an Ethernet switch in a building to support multiple customers without creating the security issues you have presented.

The main driver for presenting customers an Ethernet interface is for simplicity and cost. Ethernet is a ubiquitous and undertood by most LAN admins. T1's and T3, OC3 are not as common and troubleshooting issues can be expensive. You also need the CSU/DSU's on the customer side which is another thing that can break, plus the cost of buying them.