LwM2mNodeSenMLDecoder : Support for timestamped values

[cw8466]

Question

I have connected a client to a Leshan demo server (2.0.0-M10). That particular client adds a timestamp to the values when transmitted over SenML:
[{"n":"/3303/0/5700","t":321,"v":22.953123}]
which isn't typical of the examples provided by the specification as it doesn't specify an offset relative to a base timestamp. However, for a single value, it feels like the implementation makes sense, but the Leshan server throws this exception when it attempts to decode the SenML payload. Based on the exception, it feels like the issue is not how SenML timestamps are implemented by the client, but a lack a support altogether for SenML timestamped values by the server.
If correct, any particular reason why this is not supported? It seems like the timestamp is already parsed.

[sbernard31]

[{"n":"/3303/0/5700","t":321,"v":22.953123}]

Your client send that SenML value in a ReadResponse Correct ?

which isn't typical of the examples provided by the specification as it doesn't specify an offset relative to a base timestamp. However, for a single value, it feels like the implementation makes sense

Exactly the point, the specification is not clear at all about that.
We already have this question for Notification and my first understanding was this is not allowed by specification.
But first feedback from OMA seems to say this is allowed for notification : OpenMobileAlliance/OMA_LwM2M_for_Developers#538

The answer could be generalized to ReadResponse but this lead to some problem (as explain in OMA issue above ☝️), so I tried to get more answer about that without success ... 😞

If correct, any particular reason why this is not supported?

Adding this to Leshan API is impacting and not clear how to handle inconsistent timestamp so I would prefer to get clear answer from OMA about that...

It seems like the timestamp is already parsed.

Yep timestamp is supported for stored notification use case.

[sbernard31]

For a short term solution, note that Leshan Library allow you to provide you own Encoder/Decoder, so you can change the behavior with some code if needed.

[cw8466]

Thanks,

I confirm that it is send following a ReadResponse.

[sbernard31]

I'm curious what do you plan to do ? 🙂

[cw8466]

I'm curious what do you plan to do ? 🙂

No big deal for now. We'll avoid SenML with ReadResponse. Our usecases for timestamps is mainly through observations. But we'll also try to clarify the specification with the owner client stack as it might not be compliant to timestamp values following a ReadResponse.

[sbernard31]

Let me know what you find.

In a general way, it's difficult to me to get answer from OMA. So do not hesitate to report us that this is an issue to not get answer and maybe you owner client stack can do that too ?

(I added a new comment on related issue : OpenMobileAlliance/OMA_LwM2M_for_Developers#538)

[sbernard31]

But we'll also try to clarify the specification with the owner client stack as it might not be compliant to timestamp values following a ReadResponse.

@cw8466 do you get more information about that ?