Roman Shaposhnik 2013-09-24, 22:54
Bruno Mahé 2013-09-25, 07:54
Mark Grover 2013-09-25, 14:37
Sean Mackrory 2013-09-25, 17:54
On Wed, Sep 25, 2013 at 7:37 AM, Mark Grover <[EMAIL PROTECTED]> wrote:
> While I don't have a strong opposition to the 2 options, aren't there 2
> more options here:
> 3. We detect GROOVY_HOME and let our users download and install groovy
> themselves? We do this for Java (albeit, for licensing reasons), however,
> pulling in our own groovy just for init-hdfs.sh seems like an overkill.
> 4. Wrap the groovy script into a jar and run that jar on java instead.
> Since init-hdfs.sh is the first groovy script in Bigtop packages, I am
> inclined towards #4. If we do end up writing more and more groovy scripts,
> then we start bundling in our own groovy.
As I replied in a different part of this thread -- in this particular case
I don't actually mind simply delivering it as a java-based implementation.
That said, I think there's a great deal of benefits for scripts like this
one to be implemented in a scripting language that Bigtop users
can actually look at right there on the system and potentially extend
to suit their particular needs by simply editing the script itself.
Given that Hadoop is a JVM-based platform -- we're stuck with
JVM-based scripting languages if we want to be anywhere near
reasonable in terms of overall performance.