Re: [netmod] xpath expressions in JSON

[Robert Wilton <[e-mail]>]

On 23/10/2018 08:36, Ladislav Lhotka wrote:
> Martin Bjorklund <[e-mail]> writes:
>
>> Qin Wu <[e-mail]> wrote:
>>> -----邮件原件-----
>>> 发件人: netmod [mailto:[e-mail]] 代表 Ladislav Lhotka
>>> 发送时间: 2018年10月22日 21:12
>>> 收件人: Martin Bjorklund
>>> 抄送: [e-mail]
>>> 主题: Re: [netmod] xpath expressions in JSON
>>>
>>> On Mon, 2018-10-22 at 14:56 +0200, Martin Bjorklund wrote:
>>>> Ladislav Lhotka <[e-mail]> wrote:
>>>>> Martin Bjorklund <[e-mail]> writes:
>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> Going back to the most urgent issue, what is this WG's
>>>>>> recommendation for the subscribed-notifications draft in NETCONF
>>>>>> wrt/ their usage of
>>>>>> yang:xpath1.0 in filters?
>>>>>>
>>>>>> To summarize:
>>>>>>
>>>>>> We already have
>>>>>>
>>>>>>    o  instance-identifier in XML uses prefixes from the XML document
>>>>>>    o  instance-identifier in JSON uses module names as prefixes
>>>>>>    o  XPath in NETCONF filter uses prefixes from the XML document
>>>>>>    o  XPath in JSON query filter uses module names as prefixes
>>>>>>
>>>>>>
>>>>>> Alternative A:
>>>>>> --------------
>>>>>>
>>>>>> Use different encodings for "stream-xpath-filter" as well,
>>>>>> depending on if it is XML or JSON.
>>>>>>
>>>>>> We would do in SN:
>>>>>>
>>>>>>      o  If the node is encoded in XML, the set of namespace
>>>>>>         declarations are those in scope on the
>>>>>>         'stream-xpath-filter' leaf element.
>>>>>>
>>>>>>      o  If the node is encoded in JSON, the set of namespace
>>>>>>         declarations is the set of prefix and namespace pairs
>>>>>>         for all supported YANG modules, where the prefix is
>>>>> Is "supported" the same as "implemented", or something else?
>>>> It should be "implemented".
>>>>
>>>>>>         the YANG module name and the namespace is as defined
>>>>>>         by the "namespace" statement in the YANG module.
>>>>>>
>>>>>> Pro: the format is consistent within each encoding.
>>>>>>
>>>>>> Con: unclear how to handle other encodings.
>>>>>> Con: we keep using context-depending encodings.
>>>>>    Con: XPath expressions in JSON can get pretty long (I assume it's not
>>>>>    just an instance identifier but may contain predicates etc.). We
>>>>>    cannot use the trick with the default namespace as in YANG, so all
>>>>>    data node names will have to carry the prefix.
>>>> Yes.
>>>>
>>>>>> We could probably add that CBOR uses the same representation as JSON.
>>>>>>
>>>>>> Example in XML:
>>>>>>
>>>>>>    <stream-xpath-filter
>>>>>>        xmlns:if="urn:ietf:params:xml:ns:yang:ietf-interfaces"
>>>>>>        xmlns:ip="urn:ietf:params:xml:ns:yang:ietf-ip">
>>>>>>      /if:interfaces/if:interface/ip:ipv4
>>>>>>    </stream-xpath-filter>
>>>>>>
>>>>>> Example in JSON:
>>>>>>
>>>>>>    "stream-xpath-filter":
>>>>>>      "/ietf-interfaces:interfaces/ietf-interfaces:interface/ietf-ip:ipv4"
>>>>>>
>>>>>>
>>>>>>
>>>>>> Alternative B:
>>>>>> --------------
>>>>>>
>>>>>> Use a non-context depending encoding, with the module name as prefix.
>>>>>>
>>>>>> We would do in SN:
>>>>>>
>>>>>>      o  The set of namespace
>>>>>>         declarations is the set of prefix and namespace pairs
>>>>>>         for all supported YANG modules, where the prefix is
>>>>>>         the YANG module name and the namespace is as defined
>>>>>>         by the "namespace" statement in the YANG module.
>>>>>>
>>>>>> Pro: the format is independent from the protocol encoding
>>>>>>
>>>>>> Con: in XML, this leaf is treated differently from other XPath
>>>>>>       expressions, such as get-config filter and nacm rules.
>>>>>>
>>>>>> Example in XML:
>>>>>>
>>>>>>    <stream-xpath-filter>
>>>>>>      /ietf-interfaces:interfaces/ietf-interfaces:interface/ietf-ip:ipv4
>>>>>>    </stream-xpath-filter>
>>>>>>
>>>>>> Example in JSON:
>>>>>>
>>>>>>    "stream-xpath-filter":
>>>>>>      "/ietf-interfaces:interfaces/ietf-interfaces:interface/ietf-ip:ipv4"
>>>>>>
>>>>>>
>>>>>> My proposal is A.  I think it is more important with consistency
>>>>>> within each encoding than across encodings.
>>>>> I would suggest to consider declaring prefixes & namespaces
>>>>> explicitly in the data, as in the schema mount document. It is
>>>>> independent of encoding and the expressions can be kept short. In
>>>>> fact, one of the namespaces can be declared as default, so this use
>>>>> of XPath would then be very similar to YANG.
>>>> Ok, so this is another alternative that works today, and achieves the
>>>> goal of being encoding-independent.  It is still context-dependent
>>>> though.
>>> Yes, every module that uses XPath in data will have to deal with this. There may potentially be multiple independent prefix declarations (this is actually a con).
>>>
>>>> BTW, when used in filters, it is nice to let an unprefixed name to
>>>> match any namespace; i.e., treat "foo" as syntactic sugar for
>>>> "local-name(.) = 'foo'".  ("*:foo" is not legal...)
>>> Hmm, I think this is a bad idea because it departs even further from the original XPath semantics. Such chameleon names should IMO be pretty rare, and if they are needed, local-name() is always available.
>>>
>>> [Qin]: Agree with Lada, Referencing RFC8407, section 4.6.2, I think the below guideline is relevant.
>>> "
>>> The "local-name" function SHOULD NOT be used to reference local names
>>>     outside of the YANG module that defines the must or when expression
>>>     containing the "local-name" function.  Example of a "local-name"
>>>     function that should not be used:
>>>
>>>        /*[local-name()='foo']
>> This guideline is for must/when expressions *within* YANG modules.
>>
>> I'm talking about a different use case, namely filtering.  It is
>> pretty convenient for users to send a filter:
>>
>>    /interfaces/interface[name='eth0'/ipv4
> This is impossible if we want to call it XPath. With an explicit
> namespace/prefix declaration, for example
>
>    "namespace": [
>      {
>        "prefix": "if",
>        "uri": "urn:ietf:params:xml:ns:yang:ietf-interfaces",
>        "default": true
>      },
>      {
>        "prefix": "ip",
>        "uri": "urn:ietf:params:xml:ns:yang:ietf-ip"
>      }
>    ]
>
> it would be
>
>    /interfaces/interface[name='eth0']/ip:ipv4
>
> which is not too bad either.
This looks verbose to me.

I would still prefer using separate encodings for XML vs JSON/CBOR 
(alternative A) in the short term.

Longer term, I think that we should look at something similar to the 
JSON path encoding scheme.

Thanks,
Rob