"Eric Voit (evoit)" <[email protected]> wrote:
> > From: Juergen Schoenwaelder: Monday, December 12, 2016 6:10 PM
> > 
> > The proper action in this case may be to kill the NETCONF session.
> > 
> > It can be debated whether simply deleting state created by some other
> > system
> > (the subscription in this case) is a proper management approach since
> > it may
> > lead to undefined or erratic behaviour of other systems or to loops
> > where some
> > systems create state that others delete and so on.
> 
> Wouldn't this have implications to I2RS or OpState?

I don't know about i2rs, but it doesn't impact "opstate".

> How would an
> Operator be able to focus only on the subset of dynamic state which is
> currently impacted?
> 
> An analogy might be someone with phone and faulty video service
> connected over their broadband connection.  They get on the phone with
> an operator to troubleshoot their video, but the operator can't reset
> the video CPE session without taking out the phone service as well.

Maybe this is a good reason for using separate sessions, and not try
to do too much in one.

The problem is how you expect the client (and maybe server) to recover
after a forced termination of an ongoing request.  Note that there's
no equivalent operation to kill an ongoing <get/> for example.


/martin

_______________________________________________
netmod mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/netmod

Reply via email to