Hi, Andy,

From: Andy Bierman [mailto:[email protected]]
Sent: Monday, December 2, 2024 3:59 AM
To: Kent Watsen <[email protected]>
Cc: [email protected]
Subject: [netmod] Re: origin "system" in system-config-09



On Sun, Dec 1, 2024 at 11:34 AM Kent Watsen 
<[email protected]<mailto:kent%[email protected]>> wrote:

> I do not understand how <running> references <system>.
> How does a config leafref in <running> validate correctly unless the 
> pointed-at leaf is also in <running>?
> The same issue applies to templates if the incomplete list entry in <running> 
> has any pointed-at leafs.

This statement assumes that <intended> is subject to validation.

<running> + <system> = <intended>


Seems like some fundamental changes to NMDA., but maybe not, since now

<running> + proprietary-magic = <intended>.

My concern is for offline validation.
RFC 7950 and 8342 both say <running> MUST be valid.

There is a desire for a client to
  (1) retrieve a representation of the entire <running> datastore and
  (2) perform the YANG validation tests on the retrieved data
If <running> contains configuration that requires further transformation (cases 
defined in NMDA) and the client has the awareness that <running> alone is not 
valid (e.g., because the template needs expansion to populate the mandatory 
fields), the client then understands the magic must be done on its side before 
the YANG validation can be performed. That is also the reason I understand why 
Kent is proposing the standardization of configuration templates and inactive 
nodes.

Similarly, the client can offline merge <system> and <running> and validate the 
merging results if it has the knowledge that <running> alone contains dangling 
references to system config.

This won’t break legacy clients which still rely on the offline validation of 
<running> alone, as it’s only applicable to NMDA clients that has such an 
awareness.

There is no problem on the server because the <running> datastore is valid
if the server says so (e.g. return <ok> for edit-config or validate operations).


Kent // contributor
Andy
Best Regards,
Qiufang


_______________________________________________
netmod mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to