Hi, Martin, Please see my reply inline.
-----Original Message----- From: Martin Björklund [mailto:[email protected]] Sent: Tuesday, November 23, 2021 3:39 PM To: maqiufang (A) <[email protected]> Cc: [email protected] Subject: Re: [netmod] Should the origin="system" be required for system configurations copied/pasted into <running>? Hi, "maqiufang \(A\)" <[email protected]> wrote: > Hi, all > > There is still another issue which is about origin metadata > annotation: should the origin="system" be required for system > configurations copied/pasted into <running>? I think the question is "if a node is present both in <intended> and in <system>, which origin does it have in <operational>"? (NOTE: it doesn't matter if the value was "copy & pasted" from <system> or entered in some other way.) [Qiufang Ma] Noted, thanks. Obviously, if a leaf node is present in both, but its value differ, the origin must indicate which datastore had precedence. [Qiufang Ma] Yes, exactly. If a client tries to modify a system-defined data node(with a different value), what has been defined in <running> should take precedence over system predefined nodes if the system configuration is modifiable. Thus the origin should be indicated as "intended", there is no question about that. But suppose the node is a list entry (e.g., an interface) or a leaf with the same value. In this case, it is not clear which origin should be used. I think it would be ok to use "system" in this case. (But also perhaps it doesn't matter much). [Qiufang Ma] "Copy&paste" just refers to this case(a leaf with a same value or an ancestor node like "interface"). IMO, it should not be "intended" if the system configuration is non-modifiable(e.g., interface type or name). > Currently any system configuration explicitly declared in <running> in > order to configure its descendant nodes or maintain <running> > offline-valid will show up in <operational> with origin=intended. > The question behind this issue is whether we want a copied/pasted > system defined data node to override and take precedence over > <system>. > > The choices and some considerations of this issue received so far: > o Origin=system IS required for system configuration copied/pasted > into <running> ? I believe that "system" reflects the most accurate > source in this case. And only in this way, a server can allow a > read-only system configuration to be declared in <running>(e.g., in > order to valid > <running>) by the clients. What do you mean with "a read-only system configuration [...] be declared in <running>"? <system> is a separate datastore that clients can read, right? [Qiufang Ma] Yes, it's related to an ongoing work (https://datatracker.ietf.org/doc/html/draft-ma-netmod-with-system-00.txt). There is a read-only "system" datastore which contains system configuration non-modifiable or not. I think that non-modifiable system configuration should be allowed to write into <running>(with a same value), as far as the origin is "system". Writing a different value for non-modifiable system config in <running> should return an error. Best Regards, Qiufang Ma /martin > ? The challenge for this choice is on the server side. It MUST be > able to recognize a particular data node which explicitly defined in > <running> is actually a mirror of what is in <system>. > o Origin=system is NOT required for system configuration copied/pasted > into <running> ? Good consistency. For all configurations explicitly > defined in <running>, if they appear in <operational>, the origin > value is "intended" with no exceptions. > o Define a system-mode which is similar to with-defaults basic mode > and allow a server to advertise a particular behavior ? Does it mean > we could get the Pros from both choices? > Any other thoughts? _______________________________________________ netmod mailing list [email protected] https://www.ietf.org/mailman/listinfo/netmod
