It's fairly usual that big infrastructure systems should be regulated or publicly owned if there is no realistic competition to them. Defining that is more difficult - Railtrack presumably thought it was competing frantically with roads, after all. I propose a different way of looking at it.
What if the distinction were framed in terms of how long the assets involve last? Consider Brough Turner's graphic dealing with the useful lives of different elements in a telecoms (or isp) network. You'll see that the electronics (routers, switches, muxen etc) and optical components go obsolete within 2-3 years, but the fibre-optic (and telecoms copper as opposed to ethernet) lasts 20-30 years, and the digs last 100 years or more. In a wireless context, the kit goes obsolete quickly, the spectrum allocations on a 10-20 years timetable, and the site footprints are of indefinite life.
In a sense, design life is another way of saying money - the longer it takes to replace something, the more capital is involved, and hence the greater the monopoly power that accrues to the owner. We couldn't build another railway, and even if the alignments were available, laying track would be a monstrously huge enterprise. So whoever owns the footprint is always going to have a lot of power. Similarly, the cost of replicating the BT local loop would be impossible.
It's interesting that competition exists in fixed-line telecoms in Britain precisely at the electronic level - you are allowed to put your DSLAMs in their facilities and have your own switches, routers and such. Perhaps the rulebook for utilities should be that competition becomes more appropriate, the less time is needed to make good a bad decision? There's an analogy with education, clearly, and perhaps health - anyway, the cost of replicating NHS facilities would be mountainous.