Add support of NIDD transport for Lora and 3GPP

[sbernard31]

This aims to centralize discussion about support of NIDD transport in Leshan.

• core§6.2.1.2. Behaviour with Current Transport Binding and Modes

LoRaWAN :

• transport§6.8.4. LoRaWAN Binding
• transport§Appendix C. CoAP over LoRaWAN

CIoT :

• transport§6.8.5. CIoT Binding
• transport§Appendix D. LwM2M over 3GPP CIoT - NB-IoT and LTE-M

[sbernard31]

@Przem83, @MichalWadowskiOrange, I created a dedicated issue to talk about NIDD transport. (https://github.com/eclipse/leshan/issues/1034#issuecomment-870495461)

@madhushreegc as you already have some code which seems to run Leshan with NIDD, you may be interested (or could help) about this topic.

I link some previous issue/work/talk about that :

• https://github.com/eclipse/leshan/issues/904
• https://github.com/eclipse/leshan/issues/946

[madhushreegc]

@sbernard31 , Sure . Connector we have implemented our own to integrate with the other channels .

[sbernard31]

@Michal-Wadowski I see you start investigating on this.

I have no experience with this transport layer at all and one of my current concern about this feature is how this will be tested/testable ? (This is maybe a very naive question but does it needed any hardware or specific tool to do that ?) Do you have some inputs about that ?

[Michal-Wadowski]

I'm currently do a research with my team. At first step we want to make sure that we can just implement some kind of connector that would work with Leshan without specific IP dependencies. For now we don't know how should the final product looks like.

[boaks]

AFAIK, there is still too much stuff depending on an IP-address (InetSocketAddress). So, one of the most promising approaches is therefore mapping the "custom protocol" to something, which looks like IP.

• message oriented transfer, then use the UDP flavor.
• stream oriented transfer, then use the TCP flavor.

For both, map the "native address" into IPv4 or IPv6 addresses. Please omit multicast/broadcast addresses. With that:

• incoming messages must be converted into RawData and the "native address" is mapped into a ip-address:
• outgoing messages are passed as RawData with a destination ip-address, which must be mapped to the "native address" and the bytes must be sent with that transfer.

[sbernard31]

@Michal-Wadowski is it something Orange need ? or this is more to try to make Leshan 2.0 feature complete about LWM2M v1.1 ?

[sbernard31]

@madhushreegc, I move forward about transport layer abstraction. #1025

If you want, you could try to implement to add a NIDD with this new abstraction maybe inspiring yourself by #1312 This could help to know if we go in the right direction.

[madhushreegc]

Sure, I will try this out. I am quite busy now , later next month mid I will try.