Confusion on start time in consuming via gNMI Subscribe

[roshan-joyce-fujitsu]

Sec 6.3.3 says the following:

What does it mean the "time the record was sent"?

Shouldn't this be the time at which the record was created in the server (close to the time at which the measurement was taken).
In the case of other streams defined in TR548, a similar thing is achieved by using "token" in each record and expecting the client to keep track of the last token it has consumed; and in case of reconnection, the client specifes the token from which the server has to send again. And the server can use the information in the token to identify a specific position in the log that provides the stream.

It will be good to maintain the same concept for gNMI subscriptions also. That is, use the start time to identify a specific position in the log and start sending from there. In this case, the specified time has to be correlated to a position in the log and this position will depend on when the record was created in the log and not when the record was sent previously to any consumer.

It will be problematic for the server to keep track of when each record is sent to each consuming client.

CC: @nigel-r-davis , @bcjohnso99

[nigel-r-davis]

Hi @roshan-joyce-fujitsu, you are absolutely correct!

It should be the time the record was created/logged and not the time it was sent to the client (which is irrelevant). I think I was thinking "sent to the log", but it does not matter as it is at best ambiguous in the text! As you note, this is close to when the measurement was made, but, if there are major issues in the comms network to the device etc., the time of measurement may be very different from the time the record is created. It is assumed that the records will be created and logged in order and hence the time can be used in a similar way to a sequence number and, as you note, similar way to the the token in the compacted log.

I will taake an action to correct the text to convey this. We can review my correct text on a call.