[NGO] Re: [YANG] YANG augment-stmt

[Andy Bierman <[e-mail]>]

Martin Bjorklund wrote:
> Andy Bierman <[e-mail]> wrote:
>> Hi,
>>
>> Looking at the ABNF for the all-powerful augment statement,
>> I notice it can appear inside a list or inside an another augment,
>> anywhere a data-def-stmt can go in fact.
>>
>> Since this clause does not instantiate any data where it
>> located, but rather at the target specified by the augment-arg-str,
>> the containment for the augment seems pretty much irrelevant,
>> except that the augment could be nested within N different 'when'
>> statements, nested inside obsolete nodes, etc.
>>
>> The key in a list with an augment-stmt has no affect on the target data,
>> does it?
> 
> I'm not sure I understand this question.
> 


It seems strange to me to define an augment of 'some' other data
in a potentially different namespace, in the middle of the constructs
that define named list entries.  I'm not clear on the use-case,
which is why I cc:ed the NGO list.  We need to agree on the requirements
for data model augmentation, just like every other part of
standard NETCONF content.


>> What does it mean to have an augment inside a grouping, so that
>> it is copied everywhere a 'uses' for that grouping is specified?
>>
>> What is the use-case for augment inside augment?
> 
> The use case for augment inside anything but on the top-level is to be
> used in combination with 'uses'.  The idea is that you can do 'uses
> foo', and then in a following augment add new nodes to the foo
> grouping:
> 
>   grouping server {
>     leaf name { type string; }
>     container address {
>       leaf ip { type inet:ip-address; }
>     }
>   }
> 
>   container foo {
>     uses server;
>     augment address {
>       leaf port { type inet:port-number; } 
>     }
>   }

Is this right? Don't you mean:

    container foo {
      uses server {
        augment address {
          leaf port { type inet:port-number; }
        }
      }
    }

I understand this use-case, where the descendant-schema-nodeid
is the target, instead of an absolute-schema-nodeid.

I still don't see any use-case for augment-within-augment,
or allowing nested augments to modify data structures outside
their parent node.

> 
> Or the same thing as a reusable grouping:
> 
>   grouping server-with-port {
>     uses server;
>     augment address {
>       leaf port { type inet:port-number; } 
>     }
>   }      
>     
> This way extensions to groupings ("complex types":) can be defined.
> 
>> That means while you are defining extra data to attach to /acme:foo
>> you also define some extra data for /a:bar/b:baz?
> 
> Yes.

Why is this a feature, and what is it needed for?


> 
>> I have only seen augment examples that appear at the top-level.
>> Are there real examples of data models that actually need to
>> nest the augment-stmt anywhere and everywhere in the data model,
>> even inside the 'input' clause for an RPC method definition?
>>
>> What are the requirements for WG or vendor data model augmentation exactly?
> 
> I'm not sure I understand this question either.
> 

We should base the solution (DML) on the agreed-upon requirements.
The requirements should be based upon real use-cases.

IMO, there is a need for top-level augment clauses that name the target
with an absolute path expression, as well as nested augment clauses
that extend the data within their containment node (container, list,
choice/case-foo, rpc/input, rpc/output, notification).  It is
the DM writers choice which form to use.

You should use descendant form if 'when' clauses exist in any
ancestor nodes, instead of replicating the effective Xpath 'when'
expression for a top-level clause.

It boils down to 1 simple CLR:

    A top-level augment-stmt must use the absolute form,
    and a nested augment-stmt must use the descendant form
    of the schema-nodeid target of the augmentation.
    (A top-level augment-stmt has no context for descendants.)

The ABNF does not reflect this, or that only a choice augment
can use the case-stmt, or only an RPC augment can use the
input-stmt | output-stmt.

> 
> /martin
> 
> 

Andy



_______________________________________________
NGO mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/ngo