Hi Andy,
> NMDA never defined a "client-only" <running> datastore.
Because there’s no such thing as a client datastore? IMO, datastores only
exist with a server. What the client gets from the server seems more like a
representation.
> It requires that <running> MUST be valid without explaining how that differs
> from RFC 7950.
Again? This has been subject to a lot of discussion. I agree that RFC 8342
isn’t as clear as it could've been, but it’s pretty obvious that it meant the
<intended> is subject to validation.
> If it doesn't, then dangling leafrefs, false must-stmt, and incorrect
> when-stmt results cannot be ignored.
They are not being ignored - where do you see that? The -10 draft adds
language specific to handling the concern around dangling leafrefs.
> YANG validation rules are defined in terms of config=true or false, not
> datastores.
True, and that part doesn’t change.
> The server is not aware if a client is NMDA or non-NMDA unless it uses an
> NMDA operation.
…or if server uses a protocol (hopefully NC/RC-next) that requires NMDA-aware
clients.
> It seems like a 8342-bis would be more appropriate than yet-another optional
> add-on (system).
I like the idea of an 8342-bis, but I don’t think it should block this draft.
FWIW, I created an NMDA-next issue tracker here:
https://github.com/netmod-wg/nmda-next/issues
Kent // contributor
_______________________________________________
netmod mailing list -- [email protected]
To unsubscribe send an email to [email protected]
