Re: [YANG] insert and key attributes

Phil Shafer wrote:
> Andy Bierman writes:
>> 1) sec. 7.8.6 mentions the 'key' attribute (great, another
>>    overloaded term ;-) but this is not formally defined anywhere.
>>    The terse definition of the 'insert' attribute is not a formal
>>    definition either.
> 
> Yup, this needs some additional words.
> 
>>    In general, all the new requirements
>>    that YANG places on NETCONF agents is scattered throughout
>>    the document, and not entirely obvious.
> 
> What are you envisioning?  A section for NETCONF agent authors that
> don't want to read the entire document?
> 
>> 2) para 4 in 7.8.6 says if insert is missing, then it defaults
>>    to last.  Even in a list indexed by an integer.  This seems
>>    strange to me, to use an integral key and then ignore the actual
>>    base type, but I guess 'ordered-by-user' implies that the base type
>>    of the key has no meaning to the agent.
> 
> Exactly.  The key gives uniqueness but not order.
> 
>> 3) The last example in 7.8.7 is confusing.
>>    (There is a typo "to move a the new user 'barney'".)
>>    If 'barney' was already inserted after 'fred' (previous page)
>>    then the merge of 'barney' before 'fred' will fail in NETCONF.
>>    There is no move operation in NETCONF.  You cannot edit the
>>    value of a key.  The example would actually have no affect
>>    in NETCONF, since the <name> field was already set to 'barney'.
> 
> YANG uses "merge" w/ insert to move existing data, since not having
> a move operation is a bug.  Do we need to define our own
> "yang:operation='move'" to achieve this?
> 

I do not agree.
The WG did not want a move operation.
The set of operations is decided by consensus.
Whatever DML is used for NETCONF, it needs to work
with the protocol the way it is defined.

I think the DML should be separate from the protocol.
The YANG document should not be defining any protocol extensions or changes.

> Thanks,
>  Phil
> 
> 

Andy

_______________________________________________
YANG mailing list
YANG@ietf.org
https://www1.ietf.org/mailman/listinfo/yang