On Wed, Mar 18, 2026 at 12:30 AM James Cumming (Nokia) <james.cumming= [email protected]> wrote:
> Thanks all - Appreciate the engagement. I agree that the closest match to > the spec is to just drop the connection with no error message. I > definitely see the merits of Andy’s solution here as well but it does > indeed not conform to the spec so I suspect most clients will be unsure > what to do with it. > > Customers demanded this solution, and the default is not to add any rpc-error elements at the end. It works fine in NETCONF because data or rpc-error is expected as a child of <rpc-reply>. It doesn't work for RESTCONF since the parent node is 'output' and only its child nodes are expected. What concerns me is not that the WG was oblivious to this issue for NETCONF and then again for RESTCONF. It shows that operational experience matters. The concern is that the WG is not really capable of producing a new protocol version that incorporates new operational experience. Thanks again all > > James > > Andy > > On 17/03/2026, 05:26, "Robert Varga" <[email protected]> wrote: > > On 16/03/2026 5.13 am, James Cumming (Nokia) wrote: > > Has anyone encountered the situation where the server believes it can > > fulfil the action and starts doing so (perhaps a long running output or > > an output that returns in chunks) and subsequently encounters an error. > > In this situation the rpc-reply would have started and so an rpc-error > > wouldn’t be returned. > > Yup... so the rpc-reply starts, and somewhere during the process of > doing that we encounter a problem and realize we cannot continue. So we > do not and just terminate the session. > > Perhaps heavy-handed in that it costs us a session, but completely with > the spec, at least AFAICT. > > I would not say this is something specific to NETCONF: the same thing > happens when serving an HTTP response with chunked transfer encoding. > > Regards, > Robert > > _______________________________________________ > netconf mailing list -- [email protected] > To unsubscribe send an email to [email protected] >
_______________________________________________ netmod mailing list -- [email protected] To unsubscribe send an email to [email protected]
