RJ Atkinson allegedly wrote:
> Indeed, in the common case of an IEEE 802 LAN, the ID is used
> in IPv6 ND as an opaque index into a layer-2 bridge table for
> delivery within the final subnet.  Other link types might behave
> quite differently; I doubt that ND is in use with all link types.
> 
> In any event, bridging is not forwarding.  Even in the IEEE 802
> LAN case, one cannot even directly bridge on the Destination ID
> (since the ID might be derived from something other than the MAC
> address of the interface attached to the particular subnetwork
> specified by the Destination Locator).  So the ID really is
> just an opaque table index, even in the IEEE 802 LAN case.

(First, please ignore the message I just sent -- I'm apparently still
waking up.)

I don't think I'm parsing this right, but in any case it's clear that
the ID is used to determine the appropriate network attachment point to
deliver the packet to.  That is, the forwarding function is using the ID
as input -- to "locate" the destination.  That is quite different from
HIP as an example.

It's called an identifier because it does not determine topological
location, but in this particular case -- delivering a packet -- the only
function using this field is a forwarding function.  In other cases, for
example session control, identification functions might use it -- but
they might choose to use completely different names, for example
dynamically generated transient "identifiers".  Would this still be an
identifier if it were never used as one?

There is no clear delineation here.  Some tokens are used by forwarding
on just the last hop.  Some are used for the last few hops.  Some are
used by forwarding along the entire path.  It is difficult to say that
anything that appears in every packet is ever a pure "identifier" ...
but I assert that you don't need to.  Our goal is to support multihoming
and mobility while simultaneously solving routing and addressing scaling
problems.  To do that we need to make it possible for identification
functions to work independently of topological location.  That doesn't
mean that individual fields or names ever need to be exclusively used
for identification, although some approaches do so -- just that they can
be used for identification, independent of topology, when needed.
_______________________________________________
rrg mailing list
[email protected]
https://www.irtf.org/mailman/listinfo/rrg

Reply via email to