The main issues with this is the fact that the speeds reduce quite spectacularly as you get further from the node - while at 50m you might get ~400 to 800Mbps (

The main issues with this is the fact that the speeds reduce quite spectacularly as you get further from the node - while at 50m you might get ~400 to 800Mbps (depending on the quality of the lines), by the time you're several hundred meters away you're back to the tens of megabits... It's got to do with how the low quality telephone cables attenuate high frequencies. They're typically thin wire (0.4 or 0.64mm gauge) designed for a voice signal with a bandwidth in the kHz ranges, and then these DSL technologies use tens to hundreds of MHz, which works to some extent over shorter distances but the higher frequency you go, the more quickly attenuate over longer distances. Then you have electromagnetic interference between different subscriber's lines, which can make a huge dent in performance (literally as high as 40-50%). Vectoring is a technology that tries to model the noise and send signals to cancel it out to reduce this. It's fairly effective, but not completely. What you get is the requirement to basically have a node, fed with fibre and electricity (unless they're back-powered from houses, which was suggested for G.fast) that only services a dozen or so houses. With the cost of the capital works required, it's debatable whether it's not worth spending the ~$800-$1200 per house (this is actuals from the Australian NBN rollout) to just complete the fibre all the way, which could then in the future be upgraded to 10Gbit without any capital works (just swapping the terminating hardware in the exchange and at the customer premises - no requirement to touch any of the cabling).