Re: [Netconf] Ben Campbell's Discuss on draft-ietf-netconf-yang-patch-12: (with DISCUSS and COMMENT)

[Andy Bierman <[e-mail]>]

On Mon, Nov 7, 2016 at 6:37 PM, Ben Campbell <[e-mail]> wrote:

> On 5 Nov 2016, at 3:07, Benoit Claise wrote:
>
> [...]
>
> This text is worded badly.  I really meant "incremental", not "partial".
>>
>>> The individual edits are conceptual.  The server is supposed to validate
>>>> the datastore on the whole set, not each one.  The intent of the
>>>> sentence
>>>> was to apply all edits after validation.  e.g.
>>>>
>>>> edit-list:
>>>>   1: create /foo value=42
>>>>   2: delete /foo
>>>>   3: create /foo value=43
>>>>
>>>> (Although bad practice, allowed by YANG Patch)
>>>>
>>>> If these were real objects, like interfaces, then the system and
>>>> possibly
>>>> network could be impacted if the edits were applied 1 at a time.
>>>> Think of the edit list like a candidate datastore and the <commit>
>>>> operation.
>>>>
>>>> Maybe this sentence should just be deleted, since these are
>>>> implementation details.
>>>>
>>>
>> Either deleted, or expanded. Your corrected sentence, along with the
>> explanation, clarified it, and makes a lot of sense.
>> Let's hear from Ben.
>>
>
> I'm okay with either. I think I personally would prefer the expanded
> explanation--but I gather the point is that this is up to RESTCONF to
> explain, and hopefully RESCONF already does that.
>
>
I will add some text to YANG Patch to clarify the issue.


> I would find it unfortunate if the write yang-patch semantics varied
> materially based on the protocol that carries it; but I am fine leaving
> that to other specifications that talk about how to do yang-patch over
> whatever non-RESTCONF protocol.
>
>
The use-case I have already implemented is CoAP
https://tools.ietf.org/html/draft-ietf-core-etch-03

The CoAP PATCH and iPATCH methods are atomic, like HTTP.




>
>
>> Regards, B.
>>
>>
>>> I am fine with removing the sentence not because it is implementation
>>> detail, but because you have already clarified in Section 5 that all
>>> patches must be validated before they are applied, and that they should not
>>> be applied if one of them fails.
>>>
>>> Ben, does this deletion combined with new text around atomic behavior
>>> for other protocols address your concerns?
>>>
>>
> Yes.
>
> Thanks!
>
> Ben.
>
>

Andy



> [...]
>