Hi Qin,

> If my understanding is correct, Frank’s intention is not proposed to fall 
> back to single datastore, split tree. His concern is how Does the non-NMDA 
> client talk with NMDA compliant devices, suppose large amount of devices 
> support NMDA.

Using the original non-NMDA protocols, assuming the servers support both NMDA 
and non-NMDA.

> Does the device need to support both NMDA model and non-NMDA model?

Yes, assuming a heterogeneous mix of NMDA and non-NMDA servers. 

> Is this common case or corner case in real deployment senario.

While the industry is transitioning to NMDA, it is an expected case.  At some 
point, the IETF will obsolete non-NMDA support.

> suggestions or guidelines defined in NMDA architecture and NMDA 
> guideline(/rfc8407#section-4.23.3) seem to only assume NMDA client only talks 
> with NMDA server, non-NMDA client only talks with non-NMDA server.

True, but there’s no statement that a client or server cannot be both.  Note 
also that the NC/RC-NMDA RFCs explain how clients can discover if a server 
supports NMDA.  The intention is that the client would first try to use NMDA 
and, if not supported, fallback to non-NMDA. 

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

Reply via email to