![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
Is any of the IETF working groups dealing with the potential implications of introducing IEEE 802.1Q VLANs? Specifically, is there any effort to standardize the way routers handle 802.1Q-tagged frames? Would anybody be so kind to comment on my "speculation" presented below? (I have spent only a few days trying to get some information on the subject so forgive me if I may be sounding a little bit naïve.) Suppose an 802.1Q VLAN-compliant switch is configured with several "access ports" (i.e. ports that always receive/transmit untagged frames) and one "trunk port" (i.e. port that always receive/transmit tagged frames with multiple VLAN ID values in the tags). Let's assume the pure port-based VLAN model is used (the port based model is the only official model really covered in the most recent draft D10 of the IEEE 802.1Q proposal). Therefore, each access port belongs to a single VLAN according to its PVID. The trunk port is configured to carry data for all these VLANs so, by its nature, it belongs to multiple VLANs. Now, suppose that the trunk port is not connected to another VLAN-aware switch but to an IP router. Assuming that the router represents the latest generation of routing products I make a guess that the router will be aware of 802.1Q tagging even if the standard has not been officially approved. Does this sound like a reasonable guess? Do you know of any routers that are aware of 802.1Q tagging? If so, what does such a router do with the tagged frame? A simplistic approach would be just to: * skip the 4-byte tag (after realizing that the frame is tagged based on the Ethernet-encoded Tag Protocol Identifier), * interpret the Ethernet type field and the rest of the original MAC frame (following the tag) as if it was not tagged at all. In this simplistic approach the 802.1Q-aware router would ignore the VLAN aspect also for packets transmitted to the trunk port of the switch by supplying a null tag or no tag at all if the switch implementation allowed that. Obviously, for this arrangement to work from the IP standpoint the switch would have to be able to deliver null-tagged frames to the proper VLAN based on the destination MAC address. In other words, using IEEE 802.1Q terminology, the switch would have to implement Shared VLAN Learning as opposed to Independent VLAN Learning. The simplistic of approach described above ignores the (layer 2) VLAN aspect of the served LAN. I can imagine that a more natural approach would be to treat VLANs attached to the router via a single link of an 802.1Q switch as if they were separate segments attached to the router via separate physical ports. In this case the router would always exchange MAC frames with the switch in tagged format using VLAN IDs as "logical port" identifiers. I could not find any answers in the newest literature on VLANs or IP routing. Approaching each of the leading router manufacturers separately is time consuming and not easy because most of them show some kind of bias toward or against IEEE 802.1Q effort and try to apply a proprietary approach to combining leyer 2 switching with routing. Does anybody know how particular router manufacturers approach this issue? Thanks, Zbigniew Sufleta Systems Architect Tut Systems Pleasant Hill, CA 925-692-2354
Note Well: Messages sent to this mailing list are the opinions of the senders and do not imply endorsement by the IETF.