On quinta-feira, 29 de setembro de 2016 20:43:03 PDT Dave Thaler wrote:
> What is "this"?

The proposed filtering.

> And no we should not prevent loopback between two sockets of the same
> process.

I would say it depends.

If those two sockets are controlled by two separate stacks, they shouldn't 
filter each other's queries. Each stack is independent, so it should answer the 
each other's multicast queries.

If they are controlled by the same stack, then the stack can decide to filter 
its own loopback and simply answer the query directly from the information it 
already had. It might be easier to do that since it may have had the query 
parameters before it marshalled everything, so no demarshalling. Moreover, 
since it's coming from the same stack, it also knows about the security 
parameters and can skip the redirection to unicast DTLS port.

Or, the stack can be lazy and not do this short-circuiting. Instead, it would 
use the same code paths that it would have used for a remote query. It's less 
efficient time-wise, but it's simpler code-wise.

> I still fail to see any reason to filter within iotivity.

Let me be clear: the request needs to be answered. It just doesn't need to go 
down to the socket level.

-- 
Thiago Macieira - thiago.macieira (AT) intel.com
  Software Architect - Intel Open Source Technology Center

Reply via email to