Re: [tosca] Groups - TOSCA Language Ad-Hoc WG 2020-11-10.docx uploaded

[adam souzis <adam@souzis.com>]

> the word "resource" is even more overloaded than "object", as it tends to mean a provisionable/alloctable/creatable "thing".

Yes exactly! Is that what we are talking about?

á

On Thu, Nov 12, 2020 at 10:23 AM Tal Liron <tliron@redhat.com> wrote:

On Thu, Nov 12, 2020 at 11:30 AM adam souzis <adam@souzis.com> wrote:

Since we are talking about this on the TOSCA TC call right now, I'm not a big fan of introducing the term "object" into the spec -- I think it is too vague and the spec already has "entity" as the term as the base of its object hierarchy. Why not use the term "resource" to refer to actual, live resources and just explicitly spell out that "instances" as used in the spec today refer to logical instances in the model and not the actual resources?

Unfortunately the word "resource" is even more overloaded than "object", as it tends to mean a provisionable/alloctable/creatable "thing". But we of course support "logical" or "abstract" node types that explicitly do not become provisioned resources but are otherwise "instantiated" (or "deployed" or "installed" or "provisioned" etc.). I'm in favor of keeping the term "instantiation"âand thus "instances"âin relation to the translation into run-time reality (=orchestration), with the assumption that not all "instances" are actual "resources". Which does mean we need a word other than "instance" to refer to that intermediary ("instantiable?") model. Here's an amazing diagram:

"Template" â "Object"? â "Instance" [= in platform-local language: node, resource, instance, pod, pool, allotment, template (yes, sometimes!), vertex, session, reservation, object, etc.]

By the way, my assumption here is that we can talk about "relationship objects" in addition to "node objects", because relationships have their own intricate life (and platform-local terminology) in the run-time reality.

This is hard!