- Start Date: 2020-02-26
- HIP PR:
- Tracking Issue:
Data Credits have a fixed value of $.00001 per 24 bytes. There were original consideration to charge for Time-on-Air, but a fixed price for data usage was settled upon to simplify the experience of network users.
However, how are bytes counted? Between the LoRa Physical Header, the LoRaWAN/MAC header, and the MAC Payload, FOpts and FRMPayload there are several ways of deciding how to meter bytes.
This proposal here is to charge for FOpts and FRMPayload exclusively, but to charge 1 DC per Join uplink and 1 DC per Join-Accept downlink. In addition, any packet up or down will cost at least 1 DC.
FOpts are fifteen optional bytes used almost exclusively for MAC Commands.
FRMPayload are essentially the application data.
By defining things this way, we can keep the data calculations very simple for network users, while also protecting against simple ways to move additional data "for free".
Finally, the minimum charge of 1 DC for any packet assures that ACK packets are not free despite having potentially empty FOpt or FRMPayload fields.
Meanwhile, for network operators, there may be some confusion as "bytes on [LoRaWAN] air" do not match 1:1 bytes over the Internet backhaul. Radio packet context is communicated (frequency, spreading factor, FCnt, etc). And should the miner be colocated on the Hotspot, the message delivery protocol between Miner and the Router/Network Server will create additional overhead.
Finally, we do propose to round up DC costs; in a way this does mitigate this
point above. Therefore, if I transmit a DataUp frame with 55 bytes:
55/24.0 ~= 2.3 DCs => 3 DCs
burned on behalf of the Hotspot delivering the packet.
Striking the balance here means being fair to both sides of the market: Users and Operators.
- Helium Network Users
- Helium Network Operators
When an Operator transmits or receives a packet, they are unable to transmit and/or receive packets. As such, we must fight against exploits and charge a fee for Join and Join-Accept frames.
When a User considers the Helium Network, they know about their application and likely can estimate how large and frequent their payloads are. On the other hand, they are not expected to know of LoRa/LoRaWAN's many nuances. We believe the highest priority is to simplify their DC estimations as much as possible in order to lower the barriers of adoption.
By charging essentially 2 DC to "make the initial connection" and 1 DC per 24 bytes of application data (ie: FRMPayload) and network command data (ie: FOpts), we can resolve both points.
In addition, we charge for FOpts which is only used when issuing MAC Commands; these are generally advanced networking messages which fine tune how the device operates within the LoRaWAN stack. It is assumed that only sophisticated users would be using and thus would need to account for these. Not charging for these opens up an easy to manipulate field.
The 2 DC (1 up, 1 down) to make the connection can almost be ignored from a costing perspective, as devices can use a Sessions as long as they can preserve the session keys on both devices and network server (weeks, months, years).
Therefore, being able to simply calculate how much Application Payload is being sent per packet to estimate the cost of the Helium network will be very simple and reduce friction for potential Helium Network Users.
The rounding of DCs are a point on their own. DCs are atomic (and very low in value), so rounding up per 24 bytes used seems fair. But combined with Time-on-Air restrictions due to regulatory bodies such as FCC and ETSI, the estimated cost of data may lead to unexpected results.
For example, running in SF9, a device can only send 55 bytes in the FRMPayload.
55/24.0 ~= 2.3 DCs and I would be expected to pay 3 DCs.
But to construct a pathological example, if for some reason, I needed to
transmit a 56th byte, I would have to send a full second packet which would
cost a full third DC. Therefore, I would need to pay 4 DCs even though I may
have calculated 56/24.0 ~= 2.33
However, we believe that with an online calculator and spreadsheet resources, we can help Network Users understand and avoid this misunderstanding.
- We must establish a clear definition of how DCs are expensed. Not doing so will cause confusion amonst both Operators and Users.
This solution requires too much low-level knowledge of Network Users, although it would perhaps be the most "fair" in terms of what resource is being used
Before dicussing the remaining alternatives, it may be helpful to visualize the three nested layers of a LoRaWAN radio packet:
Radio PHY Layer
___________________________________________________
| Preamble | PHDR | PHDR_CRC |**PHYPayload**| CRC |
‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾
PHYPayload
_______________________________________________________
| MHDR | **MACPayload/JoinRequest/JoinResponse** |MIC |
‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾
MACPayload
________________________________
| **FHDR **| FPort |FRMPayload |
‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾
FHDR
__________________________________
| DevAddr | FCtrl | FCnt | FOpts |
‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾
It is possible to charge for everything sent in the PHY layer, but this would cause friction for the same reasons charging for Time-on-Air would cause friction. It requires too much low-level knowledge for users to estimate cost.
Since MHDR and MIC are fixed and this would add some fixed overhead to every packet fragment.These fields are negligible in size though and add complexity.
This would be simplest but would make JoinRequest/Response free. This would open up a simple way to use the network for free by sending payloads in the JoinRequest/JoinResponse frames.
This would make the entire FHDR free. Meanwhile, the FHDR has the FOpts field for MAC Command transport, which would be easy to exploit for free transmission of data.