CODEWITHSUNDEEP
ietf-netmod-yang
Olga
Olga

Reputation: 129

How does instance-identifier looks like in yang model?

AS far as I understand, instance-identifier type has an XPath statement which points to some node is a tree. And what's next? How does instance-identifier identify this node? How do I apply instance-identifier to the node it points to? Or do I get it totally wrong... I also don't have any example of this except those found in google like 

    leaf instance-identifier-leaf {
        type instance-identifier;
    }

Upvotes: 1

Views: 1508

Answers (1)

Paulo Gomes
Paulo Gomes

Reputation: 176

An instance-identifier can be a reference to any data node in the system. Think of it as a pointer; it doesn't contain the data itself, just a reference to it (e.g. an address) It is useful for example to represent a reference to an object that is modeled as a YANG container or list instance..

Given that YANG data can be expressed as an XML document, the way you 'point' to a specific element within that is therefore similar to 'pointing' to a specific XML element. The way you do that in XML is by using XPath, which allows to use a

Here's an example:

   container a {

      list b {
         key x;
         leaf x { type int8; }

         list c {
            key y;
            leaf y { type int8; }
         }
     }
  }

  leaf ref {
    type instance-identifier;
  }

So imagine that a real system has its datastore containing this data (for simplification, I'm using XML format, and ignoring namespaces; a real system doesn't need to keep its datastore in XML format):

<a>
  <b>
    <x>1</x>
  </b>
  <b>
    <x>5</x>
    <c>
      <y>1</y>
    </c>
    <c>
      <y>2</y>
    </c>
  </b>
  <b>
    <x>10</x>
    <c>
      <y>5</y>
    </c>
  </b>
</a>

So basically we have a bunch of list entries, some of them cascaded. If you'd represent all these nodes in xpath, you'd get a list with:

/a
/a/b[x=1]
/a/b[x=5]
/a/b[x=5]/c[y=1]
/a/b[x=5]/c[y=2]
/a/b[x=10]
/a/b[x=10]/c[y=5]

If you had an instance identifier outside this hierarchy called ref, it could take any of these xpaths as possible values, as a string value. So it contain a reference to one of those nodes; it would not contain the node itself, just a reference to it.

One final note is that it is not mandatory that the node referenced by the instance-identifier actually exists in the datastore; you can have an instance-identifier that is pointing to a non-existent node. There is a yang statement (requires-instance) that can be added as a substatement of the type statement, that allows to control whether only existing instances can be referenced, or whether non-existing ones can also be accepted.

Regarding the format of the value, note that the way the instance-identifier is represented depends on the protocol you are using. An instance identifier in NETCONF is different from one in RESTCONF (although they are very similar). You could imagine a CLI to have a custom way to represent YANG objects, for example:

  • A data node is defined with the xpath /a/b[x=5]/c[y=1]
  • In CLI, that node address is viewed as c-5-1 (just an example)
  • If you had an instance-identifier pointing to that object, its value would be the string 'c-5-1' in CLI, but in NETCONF it would still be the xpath.

In short, the format depends on the protocol you are using.

Upvotes: 1

Related Questions