Re: RPR & 802.1d
Appendix B of the 802.6-1990 spec touches on DQDB MAC Bridges. While it
does not directly apply to rpr, we might be able to use it as a starting
point. It also shows how brief such a section can be.
Also in terms of a layer service definition, it might be useful to review
chapters 3 and 4 of the 802.6 spec.
"Harry Peng"
<hpeng@nortelnetwor To: Denton Gentry <Denny.Gentry@xxxxxxxxxxx>
ks.com> cc: stds-802-rprsg@xxxxxxxxxxxxxxxxxx
Sent by: Subject: Re: RPR & 802.1d
owner-stds-802-rprs
g@xxxxxxxx
07/25/00 08:57 AM
Denny,
As for learning, each port can broadcast any new MAC address learned.
Each MAC can then operate as a half bridge for learning, learn on ingress.
However, we do have to forward all BPDU to all ports so external bridge
will disable its ports. As far as for RPR it is one LAN segment.
This means that there is no diverse routing for ring inter-connect
which I don't see it as a big problem from networking issue.
I still have many pages to pour through.....
Regards,
Harry
Denton Gentry wrote: I agreed to read through the 802.1d spec regarding
the learning
process, and how it would apply to RPR rings using destination
stripping (where the bridge might not see the packet before it
is stripped).
Overall, I agree with the sentiment that 802.1d is written to not
place requirements on the underlying MACs. I also agree that the
bridge will still function even if the destination strips the packets.
We will get more flooding of packets because the Learning process is
less effective. However, we can argue that a moderate increase in
flooding is better than having to make the ring source stripped.
I also realized something important: 802.6 (DQDB) has the same issue.
DQDB uses a pair of unidirectional links. All nodes must know the
topology, and send the packet left or right depending on its
destination address. The packet flies by its destination and "falls
off" the end. Thus, there is no single point on a DQDB net which
will see all packets. 802.1d defines how to bridge 802.6, and says
nothing about the Learning process or anything special with DQDB.
Therefore, I think we've got a strong endorsement that destination
stripping is ok.
Nonetheless, for completeness I did a strict reading of the 802.1d
spec. There are four places which discuss the learning process.
The page numbers listed below are relative to 802.1d/d17-May 25, 1998.
On #3 below, the wording is a bit troubling with respect to destination
stripping. I don't think a strict interpretation is intended here,
I'm just pointing it out.
1. pg 42, line 7 "Automatic learning of dynamic filtering
information for unicast destination addresses through
observation of source addresses of Bridged Local Area Network
traffic."
This is from the overview, and doesn't say much.
2. pg 45, line 7: "The learning process, which by observing the
source addresses of frames received on each Port, updates the
Filtering Database, conditionally on the Port state."
This wording implies that the bridge must only learn those packets
which happen to arrive at its port.
3. pg 48, line 3: "The individual MAC Entity associated with each
Bridge Port examines all frames transmitted on the LAN to
which it is attached."
The wording of this one is troubling, in that it states the
Port must examine _all_ frames on the LAN segment. Taken literally
and absolutely, this would be a problem for destination stripping.
4. pg 54, line 23: "The Learning Process observes the source
addresses of frames received on each Port and updates the
Filtering Database conditionally on the state of the receiving
Port."
This wording implies that the bridge must only learn those packets
which happen to arrive at its port.
--
Harry Peng
------------------------------------------------------------------
Dept: 1E11
Email: hpeng@xxxxxxxxxxxxxxxxxx
ESN: 39-52277
Phone 613-765-2277
Fax: 613-768-4904
Web: http://skywww/~hpeng/
-------------------------------------------------------------------