On Mon, Oct 21, 2024 at 9:18 AM David Capwell <dcapw...@apple.com> wrote:
> One thing to keep in mind is that larger clusters require you “smartly” > split the ranges else you nuke your cluster… knowing how to split requires > internal knowledge from the database which we could expose, but then we > need to expose a new public API (most likely a set of APIs) just to do > this. When you do the scheduling internal to the database you can make > “breaking” changes that improve stability into a patch fix rather than have > to wait for the next major… > As the project and its ecosystem grows we need to have a conversation on what is a public API? I do not want to derail this thread but very briefly, we should make a distinction between `project internal` private API that is exposed to Cassandra's components (which very well could run as a separate local or remote process) and public API that the rest of the world outside of the project uses. The backward compatibility expectations will be different for `project internal` private API and public API.
