> However, I still feel that we are dealing with a limitation of the YANG > framework here. I think we are dealing with a use case that was not really > foreseen in the YANG design, i.e. that we might run into data that has > instances that can indeed be authoritatively owned by a server versus others > representing more traditional configuration.
I've not followed this entire thread, but I'm a bit baffled by the discussion at this point. I thought I2RS was based on an atomic/RESTful model, rather than being transactional. What's being discussed here, though, smells a lot like a transaction based model -- one device sets some part of the "configuration" (BTW, I object to the term "configuration" in the first place, as I2RS is not dealing with configuration, but rather ephemeral state -- completely different things) in such a way that no other controller can change it. This all seems contrary to the spirit of what we're working on here... There should be priorities to understand which controller/agent has priority over the others, and there should be callbacks when your piece of state is overwritten by some other device (whether local or another controller), but there shouldn't be this concept of a "completed transaction" that we seem to be talking about here. I get the distinct feeling we're trying to boil the ocean once again -- I2RS is not going to configure the entire device (minus some trivial stuff no-one cares about). I2RS should not be trying to build a fully transactional database on each device, with locks and such. This should be fast and light, operating at the speed of any other routing protocol (on the order of milliseconds), not on the order of minutes/days/weeks/months. Can someone please explain why we're talking about this sort of "configuration?" Russ _______________________________________________ i2rs mailing list [email protected] https://www.ietf.org/mailman/listinfo/i2rs
