1. 12
  1.  

  2. 21

    For those curious why 4 or 6 specifically are a problem, since it wasn’t obvious from the thread: it appears that some routers on MPLS networks, which carry multiple types of packets on them (e.g. both Ethernet and IP packets), use the presence of a leading 4 or 6 on the packet as a quick test for whether it’s an IP packet (IP packets start with the IP version, so all currently start with either a 4 or 6). Since Ethernet packets start with the MAC address, a packet from a MAC address starting with 4 or 6 gets misclassified as an IP packet, and possibly gets the wrong thing done to it.

    1. 3

      Thank you for clarifying this

    2. 14

      Dear IEEE, please pause assigning MAC addresses that start with a 4 or a 6 for the next 6 years.

      The IEEE should do no such thing. Vendors should feel the pain of making decisions well outside of the published 802.1/802.3 standards. There is no excuse for this.

      1. 3

        Vendors might make the decisions but I think it’s us the users who feel the pain :(

        1. 2

          Pain is the name of the game with MPLS networks. Without an agreed upon safeword and an expectation of abuse, regular users shouldn’t be playing around with it. Pro-tip: Don’t let a vendor tie you up in a bunch of LSP’s unless you’re absolutely certain that’s what you want. And if that is how you choose to roll your packets, just know that the rest of the MPLS community considers it improper to whine about the bruises.

          1. 2

            Maybe I am be missing something here?

            1. Laptop manufacturer gets mac addresses assigned from IEEE
            2. User buys laptop
            3. Packets sent from laptop begin with ‘4’ or ‘6’ and get dropped by faulty hardware

            That seems like a bad situation for both the user and the manufacturer - the user is not going to blame bad networking software on hop 13 of a 25-hop lookup to Facebook, or realize that that’s the problem.

            To just say “they should suffer the pain” seems pretty misguided.

            1. 5

              My understanding is this mainly applies to ISP-level networking - your consumer-grade router/switch probably isn’t designed with MPLS in mind, and that first hop is the only one your MAC address matters for.

              As for the fix: if IEEE delays using 4/6 prefixes for six years with comparatively few such devices on the market today, what that probably means is the fix will also be delayed six years. What ICANN should probably do is allocate all their new MAC addresses under 4/6. That’ll encourage everyone to fix up their setups real quick.

              1. 2

                It is a bad situation for users, but we’re not talking about endpoint laptop users. It’s a problem for the network operator user community who has chosen to interface up with ISP provided MPLS equipment, or has decided to implement an MPLS network on their own campus.

                MPLS is a rehash of ATM networking without all of those pesky non-telco people trying to pollute the vision of bandwidth-monetized, dedicated circuit networking, like the US phone monopolies used to enjoy before customer premise equipment packet switching became the norm. It was designed with a fairly narrow deployment model in mind, under the same premise as ATM- that is it faster and easier to route packets along an initialized circuit path using a label stack than it is to packet switch using address prefix lookups. That’s the theory. In practice, it opens up a world of operational pain and tends to reward silly optimizations like this 6/4 OUI thing.

                Like ATM before it, MPLS works for a niche community of networking people, and they tend to have a level of zealotry for the technology that leads them to expect and to put up with a lot of pain from the vendors. What benefit they get, they enjoy enough that they regularly go back for more. That’s why it seems strange to the rest of the networking community for an MPLS operator to ask the IEEE to stop using 4 and 6 OUI’s. The problem is the result of a long-standing, complicit relationship between customers and vendors built upon mild, consensual abuse of the standards. It is not a problem with OUI assignment.

                I apologize for not including necessary context to frame the snarky insider comment that I made previously. I was on the team that created the Alcatel SR series IP/MPLS service routers, and the team that created the Nortel ATM Broadband Service Node. So, vendor side of that particular networking equation.