Iljitsch van Beijnum wrote:
On 28 feb 2008, at 21:08, Templin, Fred L wrote:Apologies on being slow to coming around to reading this. Still have to read further, but one question for now: is this useful for router-to-router, or is it only used for host-to-router (and maybe also host-to-host). For that matter, perhaps the definition of what is meant by "host" and "router" is somewhat soft?It's meant for each of the following cases: router-router router-host host-hostI've tried to be careful with using the words "node" and "host". Most stuff applies both to routers and hosts, but there are a few things that only hosts can do because they get to determine the size of packets that are sent, routers have no choice in the matter. I.e., the TCP MSS option makes sure a jumbo-capable host doesn't send TCP segments larger than the standard MTU to a non-jumbo-capable host, and RFC 4821 also only works on hosts.
It would be very useful to indicate when you're talking about a device or the behavior of a device, i.e.:
host vs 'device acting as a host'
It might be more useful to just call that "endpoint" and to call the
other thing a "forwarder", AND to note that whether a node is an
endpoint or a forwarder depends on its behavior to a particular packet,
NOT a property of the device itself.
The reason is that routers often act as endpoints (for control plane, e.g.); if we say "routers never do X", then a router acting as an endpoint will never do it, which would be bad. A recent case is transport mode IPsec, which ought to be required for control plane support, but is considered optional because, well, it's a "security gateway", not a "host".
Joe
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Int-area mailing list [email protected] https://www.ietf.org/mailman/listinfo/int-area
