disclaimer: Ansible creator here and I am a co-founder of AnsibleWorks.
FWIW, Infoworld didn't allow my comment on the article to be approved. I appreciate the review, but also think reducing these things down to scores is probably not the best approach, and the review was skimmed over in several places, most noticeably the end piece where it would list an advantage of one tool that was in two or three. Also, mixing in the commercial components is really more about comparing companies, which makes it confusing.
Here's what I posted:
Ultimately I think configuration systems are a bit like your favorite type of food -- and folks need to try everything to find their favorite (with a node to OpsCode) cuisine. I think that's more important than saying someone should try one or the other -- get what's going to make you productive.
To us, Puppet is a large improvement over CFEngine as a configuration management tool, and Chef contains some stylistic variations on Puppet that may make it better for many audiences (especially Ruby developers). Choose what fits your thought patterns. SaltStack is a Python implementation of the Puppet with a bit of a "different" security architecture -- to each his own.
Picking something that is secure and reliable, and that you get along with the language -- to us -- is maximally important.
Ansible's goal is a little bit different from the others -- be a general purpose workflow system, focused on apps. OS, apps, zero downtime rolling updates, should be maximally readable, infrastructure management should be a minimum of hassle, and it should be maximally secure. It can configure, but the reason you want to configure an OS is to roll out apps. And it can scale to many thousand nodes from a single box. (Alas, too much FUD about the SSH thing -- people should try accelerated mode too!). Most importantly, we believe the infrastructure that manages your infrastructure should not be something you have to manage.
So, that's what I tried to post. Anyway, I encourage everybody to experiment. I tend to think we appeal more to developers than sysadmins -- because we focus on apps, so some of the generalizations in the article I find weird, but it's also interesting to see people's takes on things.
It's always good to have attention from folks so people can find out about things -- but I think Hacker News folks should dig down beyond the generalizations, find an hour to play with everything (each), and try them in your own environment.
Likely one tool is going to resonate more -- and, I think most of us in different companies would agree -- we're fine if you think that tool is not ours, but we want it to be for the right reasons -- meaning you personally tried it :)
People coming in from a new users perspective may also be interested in Matt Jaynes's book here:
