Hi,

I have to report that at least 1 customer agrees with you about
auto-deletion.
The comment was "how are we supposed to tell there is a bug in the client
if we do not get back an error instead of silent deletion?"

A: use "must" instead of "when" and you will get an error right away.

In hindsight, perhaps NETCONF should have the equivalent of the --force
option
in many Unix commands.  e.g., only do the auto-deletion if --force is
present
otherwise return an error that auto-deletion would have occurred.


Andy


On Thu, Oct 22, 2015 at 3:45 AM, Balazs Lengyel <[email protected]
> wrote:

> Hello,
> I STRONGLY agree with Andy, Interfaces MUST work the same way.
> Autodeletion MUST work or NOT work for all interfaces (Netconf, Restconf,
> CLI, GUI, etc.) the same way. IMO it is not a protocol issue. It is part of
> the YANG definition.
>
> The whole idea behind model driven OAM is that we have one model that
> works (mostly) the same way on all interfaces. The more differences we have
> the less usable the product, the more difficult to implement.
> regards Balazs
>
> On 2015-10-21 15:07, Andy Bierman wrote:
>
>
>
> On Wed, Oct 21, 2015 at 5:46 AM, Ladislav Lhotka < <[email protected]>
> [email protected]> wrote:
>
>>
>> > On 21 Oct 2015, at 14:33, Andy Bierman < <[email protected]>
>> [email protected]> wrote:
>> >
>> > Hi,
>> >
>> > IMO we do not need lots of rules for when-stmt.
>> > They are harder to enforce than just implementing the auto-deletion.
>> >
>> > Note that auto-deletion also applies to nodes already in candidate or
>> running.
>> > It is just a derivative case to have a newly-created node deleted right
>> away.
>> > If you add node /foo it may cause node /bar and node /baz to get
>> deleted.
>> >
>> > I strongly object to treating a false when-stmt in a datastore
>> validation
>> > as an error.  This is not how YANG 1.0 works, and this is not
>> > backward-compatible.
>>
>> I think it has nothing to do with YANG (1.0 or whatever), and RFC 6020
>> correctly describes this auto-deletion behaviour for "choice" in sec. 7.9.6
>> NETCONF <edit-config> Operations. It is indeed protocol business - YANG
>> spec should just define what's valid and what isn't.
>>
>> IMO RESTCONF spec doesn't require auto-deletion.
>>
>>
>
> Our server uses the same validation engine for both protocols.
> RESTCONF does not change the behavior of YANG in any way.
> I don't see how YANG validation procedures would not apply to RESTCONF.
>
> YANG says that the node semantics apply IFF the when-stmt evaluates to
> true.
> It is up to the implementation to enforce that.  It applies to
> server-created
> nodes or nodes created via some protocol.
>
>
> Lada
>>
>
> Andy
>
>
>>
>> >
>> >
>> > Andy
>> >
>> >
>> > On Wed, Oct 21, 2015 at 5:16 AM, Balazs Lengyel <
>> <[email protected]>[email protected]> wrote:
>> > Hello Martin,
>> > I would want to codify this. My earlier proposal was:
>> >
>> > - when MUST NOT be dependent on a data node controlled by a when or
>> choice statement
>> >
>> > Notice the strong MUST NOT statement. This would simplify life greatly.
>> > regards Balazs
>> >
>> > On 2015-10-20 10:09, Martin Bjorklund wrote:
>> > I have never seen anyone trying to refer to the conditional nodes in a
>> > when expression - simply b/c it doesn't make any sense.
>> >
>> > --
>> > Balazs Lengyel                       Ericsson Hungary Ltd.
>> > Senior Specialist
>> > ECN: 831 7320
>> > Mobile: +36-70-330-7909              email:
>> <[email protected]>[email protected]
>> >
>> > _______________________________________________
>> > netmod mailing list
>> > [email protected]
>> > https://www.ietf.org/mailman/listinfo/netmod
>> >
>>
>> --
>> Ladislav Lhotka, CZ.NIC Labs
>> PGP Key ID: E74E8C0C
>>
>>
>>
>>
>>
>
> --
> Balazs Lengyel                       Ericsson Hungary Ltd.
> Senior Specialist
> ECN: 831 7320
> Mobile: +36-70-330-7909              email: [email protected]
>
>
_______________________________________________
netmod mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/netmod

Reply via email to