The current COMI draft seem to be based on the assumption that CoMI clients and 
CoMI servers share the same identifier space. In real life, centralized CoMI 
client (e.g. NMS) can implement a much larger number of objects and constrained 
CoMI clients can implement a similar size but different set of objects.

The following two example show some of the issues raised by these scenarios.

Example #1:

In this example, a sever implement YANG module A and B and a client implement 
module A and C.
Module B and C have a YANG clash but since the server don't implement these two 
modules, no rehash are created.

[cid:[email protected]]

If the client send a GET of YANG hash 727582530 (Object in Module C), the 
server will return an object in Module B instead of generating a Not Found 
error.
The client will process this Module B object assuming it's a Module C object.

Example #2:

In the previous example, we might say that CoMI clients need at least to know 
the list of modules implemented by target CoMI servers and not request objects 
not part of this list. Is this really fix the issue?

[cid:[email protected]]

In this second example, the client and the server implement the same modules, A 
and B.
However, not all objects are mandatory and the server implement the top part of 
these modules, the client implement the lower part.

The client request YANG hash 727582530 from module B and receive an object from 
module A.

The only viable solution to this issue seem to make the list of objects 
implemented by CoMI servers (YANG path, YANG hash) available to CoMI clients.
Not sure we still are in the domain of constrained devices however.

[cid:[email protected]]

Michel Veillette
System Architecture Director
Trilliant Inc.
Tel: 450-375-0556 ext. 237
[email protected]<mailto:[email protected]>
www.trilliantinc.com<http://www.trilliantinc.com/>



Attachment: oledata.mso
Description: oledata.mso

Attachment: image003.emz
Description: image003.emz

Attachment: image005.emz
Description: image005.emz

_______________________________________________
6tisch mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6tisch

Reply via email to