Stream's timestamp - blocktime versus transaction time/receive time

+1 vote
Hi,

I'm planning to use streams, so I'm performing some tests while developing the project. My current MC setting is 2 nodes, both needed to mine so that a block can not be validated without the 2 nodes being up and running (that may not be mandatory within the field of the project - but it is just the way I realised the 'issue' I may have).

Then, I have simulated 'a node crash'. Only one node is left and running, so no more block are validated, no more 'blocktime' on pending blocks added to the stream after the 'crash'. Then, I relaunch the second node, and pending blocks start again getting validated. Everything normal except that some blocks get validated at the same time, hence I have successive blocks in the stream with the exact same 'blocktime'.

Hence my question/note/remark: assuming streams can/may be used for time-based series (values submitted on a regular basis, every X minutes), it seems to me, in such situation, but also from a global perspective, that 'blocktime' can not be used in order to extract/deduce the exact time the value was submitted to the stream. It looks like to me that 'blocktime' can not reflect 'for sure' this characteristic. If one have to gather 'timereceived' within the transaction's data in order to retrieve the data submitted timestamp, how one can extract as fast as 'liststreamkeyitems' the last 100 data/values for example from a stream with their exact/effective 'submitted' time value and not the time at which it was effectively validated in the stream?

Let me know If I'm missing something here. Any thoughts welcome. Thanks.
asked Aug 31 by Christian

1 Answer

0 votes

You are right that the blocktime is not the exact time that the stream item was received, and it is not designed to be. This is a natural consequence of the consensus model of blockchains – in a peer-to-peer network there is no right answer as to the timestamp of a transaction. Each node can see the transaction at a different time, and there is no single "right answer" that can be proven to be correct.

If you want to perform genuine event timestamping, you need to embed the timestamp inside the stream item payload itself, perhaps as part of a JSON structure (if you are using MultiChain 2.0 alphas). This of course means you must rely on the transaction author to provide an accurate timestamp. In this case you can use the blocktime as a sanity check, to see that the timestamp within the transaction payload is not too far off the timestamp according to the block creator,

answered Aug 31 by MultiChain
Thanks for the answer.
I was thinking about such a solution, but it drives away the possible usage of built-in functions in order to quickly extract precise subset of data in between two given timestamps for example. Good to know.
...