ctubbsii commented on issue #3559: URL: https://github.com/apache/accumulo/issues/3559#issuecomment-1758043415
@cshannon, In general, I prefer to keep features as independent as possible. I think this is good development practice, and results in more modular code design, which is easier to maintain. It also helps with release and feature planning when things aren't so co-dependent and tightly coupled. I'm fine with targeting elasticity, though, if the benefits of doing so are substantial. On the two points @keith-turner mentioned: * The conditional mutation work isn't done yet, nor is the removal of table locks, but that's also a feature that could be done outside of the elasticity branch (possibly). It's also possible that conditional mutations can be used for this feature, without the other work in the metadata tables using conditional mutations yet, since conditional mutations are a per-write feature, not a per-table feature. I agree that there's a good chance that this may be easier or save some work by starting in the elasticity branch, because of how it relates to those things, but right now, I think this is all very speculative, as there are so many things in flight. I think think it's worth considering those things, but also making an effort to think of how it can be done independently, as the exploration of this feature proceeds. * The use cases for this may primarily exist in the elasticity branch, but that's not an argument against doing this incrementally with some progress in the main branch. If the dependency is one way, there's no reason we couldn't make some incremental changes in the main branch that is utilized more heavily later. Right now, I think it's too early to say what that might look like, since this is still at an exploratory stage. I'm okay with starting the exploration in elasticity branch, but please consider putting in some effort to make things modular and work towards incremental progress, wherever possible. If it doesn't strictly need to be implemented in the elasticity branch, I'd prefer it not be. I'd much rather make incremental progress in the main branch than dump everything into elasticity, with an all-or-nothing approach... if we can help it (we might not be able to help it... it may still make sense to bundle it all in elasticity, and if so, I'll accept that... I just don't think it's inevitable yet). -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
