Github user steveloughran commented on the issue:
https://github.com/apache/spark/pull/12004
I had something tangible: the integration tests. It's clear those aren't
wanted. Now I'm proposing something more minimal, yet still tangible for anyone
trying to build spark such that it works with object stores.
1. the POMs to set up the full dependency set, both inclusive and
exclusive. This will enable anyone who wants to build an ASF spark release to
have the consistent set of JARs with the exact point release of Hadoop they
use. Saying "oh, just drop in the ones you need" doesn't address that issue,
merely moves it downstream.
1. It actually fixes the regression wherein spark-core built with Hadoop
2.6+ drops s3n and s3 support, while still bundling jets3t JAR.
1. there's the docs to go with all of this.
1. Regarding specific things to add you want me to put something in there?
Something object store optimised like HIVE-14535 would be the obvious proposal.
I'm just feeling despondent about whether I should even bother starting
something more minimal, like a fault tolerant O(1) committer.
>I think that's only required when working with SVN and when it's
difficult for engineers to collaborate. Using the github model, it is trivial
for anybody to have a branch on github (or any other server) and have people
collaborate on that. We have had multiple instances of those with features
being developed outside, with contributions from multiple engineers from
different companies.
I don't see that as being consistent with the ASF community model. The goal
of an ASF project is to build a community around a common codebase. If instead
the process is "no, just fork into github", it doesn't appear to achieve that
goal.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
