Juergen, On Thu, Sep 25, 2014 at 08:17:54AM +0200, Juergen Schoenwaelder wrote: > Conceptually, with the shadowing approach, there is a 'merge' > operation between the config datastore and ephemeral datastores. If > the merge fails, then (as far as I recall last weeks discussion) it is > expected that a notification is emitted to the I2RS system(s) > involved.
Working through my examples to discuss this, I realized the fault in my logic: If we treat the delete operation in the local config for overlapping schemas as propagating through to the ephemeral datastore, it's effectively not a deterministic merge. It's a merge of the last party's state on top and that's not what we want. Consider the thought withdrawn. :-) > I2RS seems to seek for high transaction rates. And NETCONF people love > to keep the config datastore self-consistent. If changes to the config > datastore invalidate I2RS content, then this needs to be signalled to > I2RS clients to take action. I think it would be wrong to reject an > otherwise valid config change because it creates a conflict with > ephemeral state the config system may not even be aware of. Agreed. This does help solidify the case about the need to deal with inconsistencies introduced by changes to the underlying local config. -- Jeff _______________________________________________ i2rs mailing list [email protected] https://www.ietf.org/mailman/listinfo/i2rs
