I think this just another example of the whole ‘problem’ with Helium - apart from having to radically restructure to accommodate shortfalls & gaming of a system that is largely built on the quick-sand of notional money - the jargon & practical implementation is almost impossible to understand.
The original TTN model was simple and pure but, like almost everything like that in human history, has been taken advantage of by a minority who use the network resources disproportionately to their own commercial gain, at the expense of the majority.
So in making comparisons, it would be interesting to come up with a workable model.
It has become evident that the majority of gateway deployments are primarily for the benefit of its owner (community or commercial) who need the coverage - and as they get some side benefit from the coverage of other gateways, are inclined towards sharing their gateway on a quid pro quo basis.
It is acknowledged that there are a few totally altruistic deployments from our happy band of benevolent fanatics and those who took the leap of faith by buying a relatively expensive gateway during the bootstrapping phase. However gateways are now relatively affordable in the context of benefit accrued, so there is now little excess capital cost donated to the community network.
In a similar vein, the backhaul costs are typically negligible, leveraging existing infrastructure. For those who need to use a mobile connection there are some considerations about the recurring costs of a data plan and the proportion of own & community traffic.
Devices are deployed almost entirely for own use so any costs involved aren’t relevant to this, but see below for a sharing model.
The primary costs are in the running of an LNS. Relatively speaking, standing data is cheap to store but for the LNS to be responsive, some of the device data has be to kept in memory for instant use, even if the device only transmits once a year or indeed, never. This memory comes with a cost. There also needs to be sufficient processing power to run day to day requirements and have enough spare to cope with a reasonable unexpected surge in throughput.
I’d conclude that the most equitable funding is a small standing cost per device and a cost per uplink & downlink. It may be appropriate to credit the gateway owners a small percentage of the uplink charge, spread proportionally over the gateways that hear the uplink - as that would reduce the incentive to put a gateway in a high-traffic area that already has good coverage. Downlinks and ack’d uplinks should cost more.
A potential extension of running an accounting system may be selling data such that if an area has sensor coverage, a separate part of the stack could manage the publishing of data feeds for purchase. If someone is already collected data it makes no economic or environmental sense to deploy your own sensors. An audit mechanism of locating some verification sensors in the vicinity of the original deployment could be provided for QA.