Re: [OFF-PATH-BOF] SIP, naming, APIs and control

On Fri, 2006-07-14 at 09:43 -0700, Teemu Koponen wrote:
> On Jul 14, 2006, at 9:20, Saikat Guha wrote:
> >    _----------S---------_
> >   /                      \
> >  /                        \
> > A --- N --Internet-- M --- B
> 
> Your classification assumes M and N are not explicitly addressed  
> middleboxes, right?

The classification applies more to the signaling protocol than devices
themselves. However, the choice of the device may preclude one signaling
protocol or another. Here, N and M may or may not be explicitly
addressed depending on the nature of the device.

If N and M are UPnP/midcom/TURN/SOCKS servers etc., then they need to be
addressed explicitly, and for that, some off-path signaling needs to
happen first to discover their addresses. This off-path signaling can be
as simple as a DNS query, link-level broadcast, or a more complicated
SIP discovery/negotiation. -- Some lookup mechanism that _won't_ be used
for the actual data.

If N and M are firewall/NAT boxes, they need not be explicitly
addressed, therefore there is no such need for an off-path lookup to
discover N/M -- i.e. N and M can be discovered (invoked?) by the same
mechanism that'll be used to send data (unicast IP routing from A to B).
An off-path lookup, however, may still be needed to find B in the first
place.

Not sure if I answered your question,
-- 
Saikat

Re: [OFF-PATH-BOF] SIP, naming, APIs and control

Attachment: signature.asc