neilconway commented on issue #16622: URL: https://github.com/apache/datafusion/issues/16622#issuecomment-5646215967
@alamb High-level, an LTS in some form could definitely make sense. Coming from the more traditional database world, the fact that we have thus far gotten away with shipping API-breaking releases every 6-8 weeks is a little surprising 😅 A few questions that occur to me: 1. Six months is a pretty short support duration for an LTS release. Is that long enough that it will sufficiently ease downstream pain? 2. What should the criteria be for backporting changes to the LTS branch? Different projects use different standards here, but the more backports we do, the more reviewer time is spent checking them, and the more risk there is of regressions and disruptive behavioral changes (breaking API changes are presumably easier to avoid via tooling). Postgres has traditionally had a very conservative backport policy, for example (but they also support stable releases for ~5 years). But maybe coding agents make it more feasible to do aggressive backporting? 3. Should we consider shipping DataFusion releases less often? For example, if we shipped a new release every 3-4 months, that would reduce the frequency of breaking API changes. It might also make it easier to do press/PR about the release (more major improvements to talk about). As an example, we could do DF releases every 3-4 months and declare 1 release in 4 to be an LTS. Ultimately it would be good to understand the perspective better of what our downstream users would prefer. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
