Release plans for 0.9, shelving ansible-commander (GUI/REST project)

Hi all,

The ansible bug queue for 0.9 is getting reasonably short. I think
we should plan to release in a couple of weeks. This means feature
freeze in about a week. Thanksgiving may move that around a bit.
Almost definitely!

I need to update the changelog but there is a LOT in it and it will be
a strong release.

In other news, I have talked about my ideas around building an Ansible
GUI and REST API in the past (aka ansible-commander). I currently do
*NOT* plan to do this, due to (A) bandwidth, and (B) thinking external
inventory scripts are almost always going to fill this void first --
maybe you want to store your metadata
with your cloud instances, etc... I think that's closer to the future.
  If other peopel want to try to build this, great, I encourage that.

Some things we DO want to do are supporting multiple simultaneous
inventory files and scripts, so that if your inventory file
/etc/ansible/hosts is actually a directory, you could drop in an EC2
source and a Nova source and it would all work magically together --
along with some INI files if you wanted too (enabling conf.d style
directories as well).
It should also be possible to write inventory in plugin format to
avoid some shell calls and sharing of state/sessions between
executions. (Also, if you want to shell out to ansible-playbook,
just write a callback plugin that stores results in whatever format or
data store you like)
Any questions and so forth, let me know.

--Michael

Sounds good.

I agree with inventory plugins filling the void because for any cloud provider with an API (AWS EC2, Rackspace Cloud, etc), the always accurate inventory is the response of an API call. Then for speed, it is just a matter of caching.

Peter

It has also been suggested that for "<inventory script> --list" we
pass in the filter pattern, like "<inventory script> --list
webservers"

This would allow, at least in theory, less results to be returned for
larger host volumes.

--Michael