UUID vs. tree structured data model

amazzini commented 5 years ago

As agreed on https://wiki.opennetworking.org/display/OTCC/2019-05-21+TAPI+Meeting+Notes:

TAPI yang currently does not include any Yang 1.1 features/modules - so we can state that we use 1.0
As far as using leafref within a read-write tree to point to read-only attributes, we should use Yang 1.1 require-instance, but if we want to stick to Yang 1.0, we can use uuid instead of leafref
Decision: Replace pointer references in TopologyConstraint with UUID references for both 2.1.2 as well 2.2.0.

This decision may cause issues as far as TAPI uses YANG to define the data model, which is a tree structured data model, so naturally the first choice during the TAPI implementation would be a tree datastore. Since it’s a tree datastore, to access a subtree its always required to pass all the parent keys in order to reach the subtree. Even the TAPI YANG model expects the same while accessing all objects. Both the model driven RESTCONF API and the RPC, e.g. in "tapi-topology@2019-03-31.yang" to access a Node Edge Point from a given Topology and a given Node, the following are the RESTCONF and the RPC API structures : /tapi-common:context/tapi-topology:topology-context/topology/{uuid}/node/{uuid}/node-edge-point/{uuid} rpc get-node-edge-point-details

However, while providing the constraints during connectivity service provisioning, this hierarchy has been removed in TAPI 2.2 (which was available in TAPI 2.1) and only the leaf UUID is expected for each constraint. If the implementation is having a tree datastore, it would not be possible to access the leaf object to get its details, unless the UUIDs for each of its parent is provided.

italobusi commented 5 years ago

@amazzini : has this change be included in TAPI 2.1.2 which has been just released?

https://github.com/OpenNetworkingFoundation/TAPI/releases/tag/v2.1.2

italobusi commented 5 years ago

Since the issue is related to YANG 1.0 versus YANG 1.1, why should be change the UML model rather than the UML2YANG translation rules?

italobusi commented 5 years ago

TAPI yang currently does not include any Yang 1.1 features/modules - so we can state that we use 1.0

@karthik-sethuraman : I think it would be worthwhile mentioning this point in the Release Notes (I would start from the next release)

amazzini commented 5 months ago

This issue has been closed due to the lack of activity for more than one year. Please reopen it if follow up is necessary.

Open-Network-Models-and-Interfaces-ONMI / TAPI

UUID vs. tree structured data model #445