RE: [tosca] terminology for naming TOSCA files and templates

[Chris Lauwers <lauwers@ubicity.com>]

Hi Tal,

 

That could work, but there are TOSCA files that are neither service templates nor profiles (e.g. the files that are imported by service template files).

 

Also, if we did away with the topology_template keyword, then it would be slightly harder to validate âuniquenessâ of the keys. For example, can I have a top-level template that defines node_templates, and then import a separate TOSCA file that defines policies?

 

Thanks,

 

Chris

 

From: tosca@lists.oasis-open.org <tosca@lists.oasis-open.org> On Behalf Of Tal Liron
Sent: Monday, June 28, 2021 3:20 PM
To: Chris Lauwers <lauwers@ubicity.com>
Cc: tosca@lists.oasis-open.org
Subject: Re: [tosca] terminology for naming TOSCA files and templates

 

The subtle point is that we have two kinds of "TOSCA files" (by the way, I call them "units" internally in Puccini):

• A TOSCA file that contains a topology_template
• A TOSCA file that does not contain a topology_template (and likely contains type definitions of various kinds)

Files can import each other with the exception that a file with a topology_template cannot be imported into another file.

 

The term "service template" is meaningful, I think, because it refers to an entire collection of files (the import tree) with a topology_template at its root. A CSAR file, for example, can package an entire service template. The topology_template is one specific aspect of it.

 

Relatedly: a "profile" is also a collection of TOSCA files except that it does not have a topology_template at its root.

 

Distinguishing these things is useful. However, I agree that perhaps different terms can be used.

 

But I would like to think of this a bit differently! I've always thought that we don't need the "topology_template" keyword at all. All its sub-keywords -- node_templates, groups, policies -- could just be elevated to the main YAML level. This would reduce indentation and verbosity without any change to grammar. (By the way, this is how Cloudify does it.) Note that the same import rule should apply in this case: you cannot import a TOSCA file that contains node_templates. groups, policies, etc.

So, my suggestion ends up being the opposite of yours, Chris. Let's do away with the "topology_template" keyword and all mention of it. Thus we are left with two clear concepts:

• Service template: A TOSCA file that has node_templates, groups, policies, etc.
• Profile: any collection of TOSCA files that are not service templates.

Any file can import other files as long as they are not also service templates.

 

On Mon, Jun 28, 2021 at 4:36 PM Chris Lauwers <lauwers@ubicity.com> wrote:

I have received comments on our TOSCA specifications that suggest that our names for files and service templates are confusing:

 

• We use the name âservice templateâ to refer to a file that includes TOSCA statements. However, not all TOSCA files contain âtemplatesâ from which to create services. For example, some TOSCA files define profiles, or are included in other TOSCA files without ever defining a service.
• We also use the name âtopology templateâ. A topology template actually DOES refer to a template from which a service topology is created. Often we confuse the terms âservice templateâ and âtopology templateâ

 

I think the spec could benefit from clearer terminology in this area. How about the following:

 

• We remove the term âservice templateâ from the specification altogether.
• Instead, we use the term âTOSCA fileâ to refer to a file that contains TOSCA statements.
• We exclusively use the term âtopology templateâ to refer to a template from which services are created.

 

Thanks,

 

Chris