On quinta-feira, 15 de setembro de 2016 16:28:55 PDT Gregg Reynolds wrote:
> I confess I have not yet gotten  around to figuring out what it means for a
> node to be "routable".  could you say a few words on that or point me to
> some docs?  and as long as I have your attention: what the relation between
> client/server on the one hand, and EP/GW on the other?  e.g. is a server
> that is not also a client always an EP?

I'm not sure how far the GW implementation goes to implementing the Resource 
Directory spec. Basically, it discovers devices on another non-routable 
network (usually, between an IP and a non-IP ones) and forwards the packets 
from one network to the other. So when a GW sees a discovery, it will answer 
with discovered resources from the other network. Later, when it sees a 
unicast packet targetting one of those resources it had announced, it will 
forward to the target endpoint and from it back to the original client.

Whether the GW will pre-discover resources or whether it will do on-demand 
discovery is an implementation detail.

One thing I am not sure of is whether we've fixed the requirement that the 
endpoints pass extra information. That's unacceptable: the GW has work 
transparently for each and every client, including baseline implementations of 
the OCF protocol. If the GW protocol requires that, then it's a design flaw and 
needs to be fixed.

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

Reply via email to