3 ms·

> Dual-stack is a terrible curse and a hack. Odd. Is the IPX + IP dual stack a terrible curse and a hack? [0] IP + Ethernet? [1] [0] If you think this isn't

> Dual-stack is a terrible curse and a hack. Odd. Is the IPX + IP dual stack a terrible curse and a hack? [0] IP + Ethernet? [1] [0] If you think this isn't real, you're not old enough. [1] Given that your focus is home networks, other than "legacy baggage" there's zero reason that switches and IP implementations couldn't eliminate Ethernet addresses and use IP as their lowest logical layer. Switches already have to maintain a "MAC address -> switch port" lookup table. Converting that to an "IP address -> switch port" lookup table is easy. With the removal of Ethernet addresses, the need for ARP is eliminated, so there's no need for hosts to remember MAC addresses. Things are much simpler without all that legacy baggage, don't you agree?

ButlerianJihad 2mo ago
Eliminating MAC addresses is a novel idea. I have not seen it credibly suggested before. Surely some functionality would be lost? Surely you'd at least need to redesign all Ethernet controllers around it? Anyway, IPX, and Ethernet MAC, are different protocols with unique aims. I guess IPX was routable. But it was proprietary; it was Novell's baby, and it was serving a particular purpose for their software and network architecture. There was no sin in running it parallel to IP stacks. The sin in dual-stack IPv4 + IPv6 is that they're the same protocol. There is nothing different in their aims, their reach, their utility or their transit networks. They're all serving the same Internet purposes, and the same protocols are both on top and below them, layer-wise. Same old TCP, UDP, QUIC, etc. So, the only reason 4 and 6 coexist is transitional support. The only reason 4 still needs support and dual-stacking is because of a lack of 6 support. When/if IPv6 is wholeheartedly embraced, then dual-stacking is obsolete. I'd like to reiterate that dual-stacking is an administrative pain and possibly the source of some really crazy connectivity bugs. A dual-stack system is unnecessarily complex. Eliminating Ethernet and MAC addresses is a revolutionary idea that would simplify things. But today, it is realistic to eliminate IPv4 just by widening support for v6. And that's something we should be working towards.
- simoncion 2mo ago
  > I'd like to reiterate that dual-stacking is an administrative pain... This professional administrator disagrees. > The sin in dual-stack IPv4 + IPv6 is that they're the same protocol. There is nothing different in their aims, their reach, their utility or their transit networks. That same commentary applies to IP + IPX. Your understanding of IPX is poor. This was about the depth of analysis I was expecting. Pity. > Eliminating MAC addresses is a novel idea. I have not seen it credibly suggested before. Surely some functionality would be lost? No more than eliminating IPv4 addresses in the far future, and far less if that elimination was done today. Go read up. It'll probably take you a few weeks to get up to speed. > But today, it is realistic to eliminate IPv4 just by widening support for v6. lol.
- wmf 2mo ago
  When/if IPv6 is wholeheartedly embraced, then dual-stacking is obsolete. ...dual-stacking is an administrative pain and possibly the source of some really crazy connectivity bugs. A dual-stack system is unnecessarily complex. I don't think anybody really disagrees with this but it's kind of pointless to mention because there is no alternative. IPv8 would also require a period of dual-stacking. T-Mobile is already running 99% single-stack IPv6-only so I would say the transition works; it's just a matter of doing it.

ButlerianJihad 2mo ago

Eliminating MAC addresses is a novel idea. I have not seen it credibly suggested before. Surely some functionality would be lost? Surely you'd at least need to redesign all Ethernet controllers around it? Anyway, IPX, and Ethernet MAC, are different protocols with unique aims. I guess IPX was routable. But it was proprietary; it was Novell's baby, and it was serving a particular purpose for their software and network architecture. There was no sin in running it parallel to IP stacks. The sin in dual-stack IPv4 + IPv6 is that they're the same protocol. There is nothing different in their aims, their reach, their utility or their transit networks. They're all serving the same Internet purposes, and the same protocols are both on top and below them, layer-wise. Same old TCP, UDP, QUIC, etc. So, the only reason 4 and 6 coexist is transitional support. The only reason 4 still needs support and dual-stacking is because of a lack of 6 support. When/if IPv6 is wholeheartedly embraced, then dual-stacking is obsolete. I'd like to reiterate that dual-stacking is an administrative pain and possibly the source of some really crazy connectivity bugs. A dual-stack system is unnecessarily complex. Eliminating Ethernet and MAC addresses is a revolutionary idea that would simplify things. But today, it is realistic to eliminate IPv4 just by widening support for v6. And that's something we should be working towards.

simoncion 2mo ago
> I'd like to reiterate that dual-stacking is an administrative pain... This professional administrator disagrees. > The sin in dual-stack IPv4 + IPv6 is that they're the same protocol. There is nothing different in their aims, their reach, their utility or their transit networks. That same commentary applies to IP + IPX. Your understanding of IPX is poor. This was about the depth of analysis I was expecting. Pity. > Eliminating MAC addresses is a novel idea. I have not seen it credibly suggested before. Surely some functionality would be lost? No more than eliminating IPv4 addresses in the far future, and far less if that elimination was done today. Go read up. It'll probably take you a few weeks to get up to speed. > But today, it is realistic to eliminate IPv4 just by widening support for v6. lol.
wmf 2mo ago
When/if IPv6 is wholeheartedly embraced, then dual-stacking is obsolete. ...dual-stacking is an administrative pain and possibly the source of some really crazy connectivity bugs. A dual-stack system is unnecessarily complex. I don't think anybody really disagrees with this but it's kind of pointless to mention because there is no alternative. IPv8 would also require a period of dual-stacking. T-Mobile is already running 99% single-stack IPv6-only so I would say the transition works; it's just a matter of doing it.