Jeremy Kong

Argument for this-is-not-worse-than-transactions2 in terms of correctness: - The main failure mode with `acceptStagingReadsOnVersionThree` is that if we read `(V, STAGING)` but the other two nodes have `(W, STAGING)` or...

Thanks @gsheasby. I'll cut an RC

This needs a develop version, which I'll work on shortly.

Encountered this in https://github.com/palantir/atlasdb/pull/3461 - manually hitting the endpoints seemed to work, but was getting failures (and yeah we use a docker in docker setup)

> Could probably also separate out all the postfiltering logic into its own PostFilter class Yep, there's actually an internal project ongoing to slice this into many parts (see #7000...

I noticed that when I run the publish task on atlasdb proxy I get ``` - Variant shadowRuntimeElements: - contains dependencies that cannot be represented in a published pom file....

Thanks for the PR! This looks good though we were planning on doing similar in the key-value-snapshot-reader chain (#7076 and following PRs). Will hang on to this for now, though...

> Not sure if we can know there are no rows to delete though - either would have to be informed by client or understood at read-time Yep, exactly. >...

Ah, I was concerned (maybe too concerned?) about breaking users who might rely on existing whitespace, though I can't think of a use case for that offhand. If you think...

@CVdV-au Makes sense, thanks. Assuming we do put in the work for that, I'm trying to get a feel for if/how much the end state would be better: my intuition...