I had a long argument with my peers on the difference between the three ways of doing canary releases. I know there are more but the current set in question are these three below
- Config service
- Feature service
- AB testing
Following are the doubts
- Can someone help me understand why a config service can/cannot be used as a feature service?
- if someone pre-fixes the userIDs participating in the A and B cohorts, will it still qualify as an AB test? My understanding has been that the cohorts must be randomly chosen to remove bias while I am being told that it’s just one of the strategies
To some, it may seem all of this is just nomenclature but names are important. They anchor concepts.
It would also be great if someone can refer to me the right literature so that me and my peers can read it and improve our understanding.
