carp.core-kotlin
carp.core-kotlin copied to clipboard
Make sequence ID verification optional?
Sequence IDs in data streams have a valid reason for being there:
- detecting dropped data
- allowing the backend to know no more data with older timestamps will be uploaded (unless sync points change, which is a separate topic)
However, to simplify client implementations, there has been a request to not have to be bothered with tracking sequence IDs. A specific infrastructure can therefore decide to omit sequence ID constraints. But, this inhibits using the current DataStreamSequence and DataStreamBatch objects, since these enforce these constraints.
Could there be an easy way for infrastructures to disable sequence ID constraints? Is this at all desirable?
Could there be an easy way for infrastructures to disable sequence ID constraints? Is this at all desirable?
YES! This is highly desirable - this whole sequence ID issue is causing us many problems in CAWS right now.
Other than redeployment needing the last IDs, are you running into other issues?