----- Original Message -----
> From: "Andy Parker" <a...@puppetlabs.com>
> To: puppet-dev@googlegroups.com
> Sent: Friday, September 7, 2012 5:53:34 PM
> Subject: Re: [Puppet-dev] Hiera-puppet with Puppet 3 - FakeScope
>
> On Fri, Sep 7, 2012 at 9:10 AM, R.I.Pienaar <r...@devco.net> wrote:
> >
> > but what if you had something like this:
> >
> > class example::data {
> > case $::ofamily {
> > "RedHat": { $service = "httpd" }
> > "Debian": { $service = "apache2" }
> > default: { fail("Please specify a service name using hiera")
> > }
> > }
> > }
> >
> > Here we clearly model different behaviours depending on the context
> > and
> > clearly state what we do not support via default but still alow the
> > module user to then decide they would provide this override data
> > through whatever hierarchy make sense for their site, ENC or just
> > hardcoding it.
> >
>
> The data bindings can handle that by having the Debian and Redhat
> cases in hiera (all of the facts are available to the data binding
> service) and then the class would have:
>
> class example($service = "UNKNOWN") {
> if ($service == "UNKNOWN") {
> fail("Please specify a service name")
> }
> }
>
> I don't think this is quite as clear as the data class form, since
> there is now no single place to look to understand the behavior.
this assumes everyone has operating system hierarchies, the data class
example specifically makes the class multi os aware regardless of whatever
hierarchies users have provided.
In your example you'd either need to have a similar case in the example
class - thus coupling the data processing into the logic class - or
force everyone to have OS hierarchies AND supplying all the data you
need.
The point of a in-puppet hierarchy is that it's there by default and
works by default as a last resort in the case that a user doesn't provide
the data.
It's also a good crutch in the case where people want to do crazy logic
on their data using complex if/case/math which they couldn't do using the
yaml/json/whatever data sources as those clearly can't contain any kind
of logic other than that provided by the fact based selection which is
roughly equivalent to a case statement
consider:
class exim::data {
$deliver_queue_load_max = $cpu_cores * 1.2
}
Here we want exim to not deliver mail when the server is too loaded
but what defines 'too loaded' depends on the amount of cores the machine
has. It would be pretty painful to express this in pure-data hiera
as you'd need to dedicate an entire hierarchy to the amount of cores
present, quite a pain
I'd think the only sane alternative is:
class example($service=$example::data::service) {...}
which roughly achieves the same as the hiera-puppet backend in this
specific example, not particularly nice though I'd think we can do
better
>
> >> separate data from code. The data bindings do this by extending
> >> the
> >> language and providing an extension point in puppet where you can
> >> plug in your own data lookup services. The puppet backend was a
> >> hack that
> >> depended on knowledge about puppet that leaked through to the
> >> implementation of the backend, which was never supposed to be
> >> available. We made the decision to not continue leaking this
> >> information in the data bindings code so that it would be less
> >> tied
> >> to the internals of puppet. That had the effect of cutting off the
> >> puppet backend, which we did not consider a problem since we
> >> wanted to
> >> promote the pure data bindings going forward.
> >
> > I think the puppet backend far from being a hack was actually a
> > major
> > win for module authors and our ability to create reusable modules
> > that
> > work out of the box.
> >
>
> True. And the data bindings are trying to take a next step in that
> direction.
>
> > I think the 'hack' aspect comes from the fact that puppet doesn't
> > cleanly
> > expose some of the data this backend would need hence the fake
> > scope thing
> > but conceptually I think if it was realised better this backend
> > still has
> > solid reason to exist and function
> >
> > Either way, if we are not going to support a puppet backend the
> > pull
> > request 1111 needs to remove that backend.
> >
>
> I'm not sure if we want to remove it yet, since it still works and a
> lot of people are still using it. We would probably deprecate it
> first. I would only want to do that if there was enough consensus
> that the data bindings provide enough of an alternative solution.
>
> So we are going to continue supporting the puppet backend via the
> hiera functions. The data bindings will "work" with it in the sense
> that it doesn't blow up (as it did the first time we tried a setup
> with it enabled). At some point during the 3.x cycle it will probably
> be deprecated as it becomes clear that its use cases are covered by
> the data bindings, which might take some changes to the data
> bindings.
OK
--
You received this message because you are subscribed to the Google Groups
"Puppet Developers" group.
To post to this group, send email to puppet-dev@googlegroups.com.
To unsubscribe from this group, send email to
puppet-dev+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/puppet-dev?hl=en.
