If a pure-java re-write of winrm4j will make our life significantly
easier for OSGi'fying Apache Brooklyn, then yes.
But if it's going to be a lot more effort to rewrite it, then we should
live with it for now.
---
My gut feel is that a rewrite won't be that hard, given we have the
reference implementation in Python.
We also have pretty good live-test coverage for various different
scripts (I'm copying what we have in Brooklyn
WinrmMachineLocationLiveTest into the winrm4j repo).
Aled
p.s. we are trying to get the OSGi stuff done within the next couple of
weeks, to include it in a 0.9.0 release.
On 12/10/2015 17:25, Valentin Aitken wrote:
+1
However exactly because it is just 822 it is really not a problem that
it is a python code.
We are using it intensively and we didn't hit any problems still.
So I vote to do it but later.
Valentin
On 10/12/2015 07:20 PM, Richard Downer wrote:
// new subject, was [PROPOSAL] Split brooklyn-core during OSGification
The core code of PyWinRM (i.e. excluding tests) is 822 lines of code.
I know that Java line counts tend to explode, but I'm thinking that
maybe creating a Java implementation of the WinRM client is not such a
horrible task as I thought it perhaps was. Doing this would mean that
we could drop the Jython dependency. It's a disproportionately large
dependency for the value it gives us.
https://github.com/diyan/pywinrm/tree/master/winrm
Thoughts?
On 12 October 2015 at 17:14, Alex Heneveld
<[email protected]> wrote:
We should move winrm/jython to a new project in any case. Many
users will
want to exclude that bundle.
I suggest software/winrm/ folder brooklyn-software-winrm project.
--A
On 12/10/2015 17:59, Hadrian Zbarcea wrote:
I think we'll need to split the core in multiple parts actually. I am
having a bit of trouble myself with winrm4j. The complication is
that it
depends on jython, which does not provide an OSGi bundle either. I
am not
sure for instance if winrm4j is always necessary or it could be
another
core-ish bundle.
A second thought is that I don't know what the impact on current users
would be to move the current use of the OSGi framework in a
different jar.
One possibility is to split the core into multiple jar, following a
convention we can agree upon (brooklyn-rt-* would be a common
convention),
but then let what Cipi called brooklyn-rt-felix, keep the current name
brooklyn-core.
Cheers,
Hadrian
