[Hawkular-dev] Inventory configuration definitions

[Lukas Krejci]

Hi all,

There is a pending PR on inventory to add support for checking configurations 
against a schema. The configs are directly mappable to JSON and the schemas 
are actual JSON schemas.

Now to create a resource with configuration is a two-step process. First one 
creates a resource without configuration and then adds a configuration to it. 
When the PR is merged the configuration will be checked against the schema 
(stored at the resource type) but that check will only happen when the config 
is created (or updated), not when the resource itself is created.

Today, Heiko asked if a URL resource could be made required to have a valid 
URL. Given the above, the answer is yes, but the URL resource can exist 
without any URL, too, until someone assigns one to it (at which point it can 
be ensured that it is a valid URL using some regex magic).

My question is, do we actually want to support Heiko's scenario (i.e. disallow 
resource creation without also supplying valid config)? In my mind, one of the 
reasons for bothering with Hawkular and abandoning RHQ was to ease up on the 
strict rules we had there about resource hierarchies and resource creation. 
Certainly, those restrictions and rules made sense, but at the same time 
forced a rigid and non-modifiable workflow when working with RHQ.

IMHO, there should be an understanding of an "incomplete" resource - i.e. one 
that doesn't have all the data with it that is should have - either because 
that piece of data hasn't trickled down to the inventory or because the user 
actually didn't decide yet what the resource will end up being. Having the 
support for incomplete resources also makes the future support for changing 
the resources' types easier - it could be done in steps.

Do you see strong reasons for requiring valid config at resource creation time 
or do you think that current workflow is "workable" and my reasons for doing 
it that way at least partially valid?

Thanks for input,

Lukas