[Puppet-dev] Re: [augeas-devel] Re: Augeas and Puppet

[David Lutterkort]

On Fri, 2008-08-01 at 16:00 +0200, David Schmitt wrote:
> I have to admit that I don't get the fine points of the discussion 
> there. Is there really a use case for "mv" at all? Especially in the 
> puppet context, where there is much less programming and much more 
> specification?

You're right that there is a big difference between how Augeas thinks
about the world (action-oriented) and how Puppet thinks about the world
(state-oriented)

Bridging that gap is one of the challenges of coming up with an Augeas
type for Puppet. If you want to be very strict, you'd have to always
describe the (sub)tree you want Augeas to see in your Pupet type; for
example, you'd describe one line in /etc/hosts by saying 'remove the
subtree for 192.168.0.1 and then set it to have the following nodes with
the following values' - it is then the job of the Augeas type (or Augeas
itself) to make sure that that only leads to a file change if there was
actually a net change in the tree.

This pretty much gets you to the same problems as what the file type has
to address - what Augeas does essentially is let you think about the
contents of config files as files in a filesystem (where the 'files' in
Augeas' tree correspond to individual entries in the underlying config
file)

David