Thanks for the input Benjamin. Sounds like you've come to a lot of the same conclusions I have. I'm certainly keen on fixing up MV's and I don't really see a way we could avoid it, as I know they are already widely being used in production. I think we would have had a much easier time if we went with a basic implementation (append only) first, but y'know, hindsight. Unfortunately I'd say we're kind of stuck with fixing what we've got or have a really angry userbase that jumps ship.
*What I miss is a central consensus about "MV business rules" + a central > set of proofs and or tests that support these rules and proof or falsify > assumptions in a reproducible way.* > >From what I gathered from JIRA the goals in my original post are the ones outlined during initial development of MV's. The general design and goals were also documented here <https://docs.google.com/document/d/1sK96wsE3uwFqzrLQju_spya6rOTxojOKR9N7W-rPwdw/edit#heading=h.c88v9p3byo75>, however doesn't completely cover the current state of MV's. I'm with you that we certainly need a set of proofs/tests to support these rules. At the moment a lot of the open tickets have patches that contribute good tests that cover many cases however we're almost kind of defining rules as we go (granted it is difficult when we need to test every possible write you could make in Cassandra). In regards to your "tickler", a colleague has been working on something similar however we haven't deemed it quite production ready yet so we haven't released it to the public. It may be useful to compare notes if you're interested!