Requested revision
Standard: | IEEE Std 802.1AB-2016 | Clause: | 6.7 |
Clause title: | Systems with multiple LLDP Agents |
Rationale for revision
Quoting from a sentence in the first paragraph of 802.1AB-2016 clause 6.7,
"It follows that each LLDP agent makes use of a unique MSAP, and that the agent can be uniquely identified by the receiving agent using the MSAP’s identifier as specified in 6.2." Figure 6-2 verifies this sentence. If a system implements multiple LLDP instances on a port, each one requires its own MSAP, and thus, its own MSAP identifier, which is a MAC address.
Quoting from the IEEE Standards Association's "Guidelines for Use of EUI, OUI, and CID" (https://standards.ieee.org/content/dam/ieee-standards/standards/web/documents/tutorials/eui.pdf):
"With the exception of such protocol identifiers, EUI-48 identifiers are still intended to identify items of real physical equipment or parts of such equipment, such as separable subsystems or individually addressable network ports. The expected use should not exceed one EUI-48 identifier per hardware subsystem, or at most a very low number of EUI-48 identifiers per physical instance of such equipment (e.g., groups of ports as in IEEE Std 802.1AX, for link aggregation). Allocation of a single EUI-48 identifier to identify or permit addressing of a fixed and permanent function associated with a real item of physical equipment occurs for the lifetime of that equipment or an indefinite period of use."
That is, the RAC prohibits (very appropriately!) using MAC addresses to identify software constructs, which multiple LLDP instances certainly are.
I do not believe that there is any justification whatsoever for this requirement for multiple MAC addresses for multiple LLDP instances, as the LLDP makes no use whatever of the source MAC address in an LLDP frame. One could, just barely, justify the extra MAC addresses required for multiple agents on one physical port using the "or at most a very low number" of addresses phrase in the guidelines on the grounds that you expect to implement at most two or three agents on one port at different reaches. But, if unicast addresses are used as destinations for agents, then the number of MAC addresses required per port can be as large as the number of other devices that could be attached to the bridges LAN, and that is absurd. For that matter, even for the two agents with two reaches case, from where does this extra MAC address come?
Proposed text
I have no proposed revision text at this time. The problem is a fundamental confusion in this document between LSAP and MSAP. Figure 6-2 is fine in describing the normal case, but should include the case of LLDP Agents and their individual LSAPs sharing a single MSAP. Each instance receives all incoming frames and differentiates the frames it wants to receive by destination MAC address. All transmit using the same MSAP and thus the same source MAC address. A list of all confused uses of MSAP and LSAP is needed, as well as, at least, text for 6.7.
Note that this affects the new work on P802.1ABdh, as it makes use of the different MAC addresses to differentiate which data base is being queried or supplied in extension PDUs.
Impact on existing networks
Unknown for sure. It's hard to imagine that people are generating separate MAC addresses per LLDP instance, but I suppose it's possible.
Originator
Name: | Norman Finn | Email: | nfinn@nfinnconsulting.com |
Affiliation: | Huawei | ||
Submitted: | 2020-11-23 |