Github user srowen commented on the pull request:
https://github.com/apache/spark/pull/5085#issuecomment-84141192
There are two questions here.
"Should assembly file names be required to contain `hadoop`?"
Yes argument: it's the current behavior and might avoid picking up other
random assembly-like JARs
No argument: might help people pick up a customized assembly JAR more
naturally.
I personally think it doesn't matter much, and all else equal would leave
it, considering the discussion here. Is there an end-user problem that hasn't
been voiced yet?
"Should the launcher .jar be separate from the assembly?"
Yes argument: slightly faster startup. Maybe a desirable decoupling? it's
how it works now.
No argument: deployment now means managing two files
I personally don't think the startup time difference matters. Right? This
is a deployment question. It's not worse than when one had to ship around both
scripts and an assembly, so I find this the same as before. That doesn't mean
there isn't an argument for one uber jar.
Is that it then? question is: should the launcher be a part of the uber jar
to simplify deployment over the current and past state of things, at the cost
of a little startup time hit? I could see that one go either way.
---
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]
