webwat: Government doesnt invest in peering unless a university decides to run a little peering business on the side. The peering switch requires users to connect, so of course port fees and related fibre services generate revenue.
The subscriber density issue has no effect on the economics of actual peering or how to get fibre to an population area, and I am sure most ISP use Ethernet equipment so its at least as fast as the peering switches. Most ISPs would have as much local bandwidth as any NZ peering exchange can cope with. Main problem is how much it costs to buy nationwide backhaul from residential subscribers through Telecom's network, and how many other residential customers would start complaining about congestion in their area or on their ISPs main Telecom link.
There is also the question of how much bandwidth can APE handle. Its a big switch but I dont think especially fast. Maybe someone could comment on how much close APE would be to having serious congestion in the case that local IPTV suddenly attracted several thousand residential users. Or is it already congested? That would explain why ny nationwide traffic is so slow already!
I'm sure someone from Citylink can chime in; but APE and WIX are effectively virtualised services (VLAN) on their public Ethernet fabrics. Given the nature of their network is commercial, I suspect they're unlikely to congest it. In the face of massive demand there's the possibility for momentary congestion or while they react with upgrades - but as you point out the constraint is getting the bandwidth to the end subscriber.
Regarding the subscriber density, it does have a small bearing - if you have significant density in an area you may want to hand that traffic off to peers as close to your BRAS/BNG as possible, rather than hauling it to Auckland or WIX. This has been Telecom's argument for introducing their 29 points of interconnect for lfully localised peering. Of course, Telecom's approach has drawbacks - 29 locations builds a lot of cost for peering partners to reach them all (or they must buy backhaul over TNZ's network to reach them).
Of course, ISPs which only have a single POP for their BRAS/BNG aren't necessarily interested in regionalised/localised peering - it's easiest for them to haul it all to Auckland (or Christchurch or where-ever they are).
While I'm typing this I thought I'd close off on the below as well.
exportgoldman: Why not have the government fund what Telecom proposed last year, that a series of 13(?) local peering exchanges are placed around the country and anyone which brings there own cable to that location can peer for free.
Free national traffic really fell out of the tree when TelstraClear decided to make a gob of money from it, and now local traffic trombones through Australia for local NZ content.
How is THAT cheaper than local peering...
Telecom's proposal was for 29 locations - to interconnect with their subscriber base only. While a nice step forwards by Telecom (and completely flies in the face of TelstraClear's stance), it has drawbacks.
I'm not sure the Government needs to fund peering - there are already multiple IXs in the country, easily accessible and modestly priced, but it certainly needs to encourage the two biggies to participate. Of course, as I mentioned previously, it could be in the "peering ISPs" favor to actually develop services that make use of their highly interconnected nature and start to attract revenue away from the biggies -- i.e. actually innovate for once. This has certainly happened in Australia, with the independents vs. The Gang Of Four.
As for the last comment - it should be vary rare that domestic traffic within New Zealand actually trombones Australia (or any other country) and is usually a configuration anomoly. Virtually all ISPs in New Zealand do have connectivity to TelstraClear or Telecom via their paid circuits, it just may not be cheaper for them to use this on a direct $/Mbit basis (but the other benefits, like it not being so latent, are worth it).