IPv4 will be around forever. It'll just get more expensive.
The only thing I can see replacing it is if something I'll call "easyip" came around - something even more convenient than ipv4 for networks of under, say a few hundred devices
Only if you’re not the one who’s responsible for designing and managing cidr allocations
Again, the only replacement I can foresee is for an address schema that either doesn't pretend to even attempt global space or has a UUID scheme where network addresses are immutable, globally unique and inviolable.
This would potentially simplify an iot world if the addresses are known, globally unique and static.
You mean like IPv6 ULA? =)
No.
Currently when you get an iot device you usually connect to it through the smartphone, tell it your Wi-Fi hotspot, give it the password and have it go online.
Instead there could be a flow of going to the routers interface, select "scan new qr code" and you add the device that way - more realistically this would be done through the "myspectrum" or equivalent app as the isp usually issues the modem/hotspot for the majority of consumers.
A number of things would have to change to make that possible but it's a much easier setup flow and it likely requires less complex software on the device itself.
If you want really simple solution on device itself, just let it NDP for router info and drop any incoming SYN that's not already in the conntrack table or part of the ULA space. Boom - done.
(The only answer that makes sense here is supporting legacy devices, which will probably be a problem we'll have to deal with for the rest of our lives. Thankfully I don't appear to have any such devices on my network!)
Not that I'm advocating anyone use that particular subnet of fc00::/7 for their local network.
I dual-stack for my internet gateway.
IPv6 networks are simpler. They do many things automatically that have to be entered manually. The big one is don't have to worry about subnet size or subnet masks; all subnets are /64 and effectively infinite. This doesn't matter for home network but it is big advantage for corporate networks.
All I really need is to be able to accommodate IPv6 on the internet side, which I can currently do.
Alas, being on the global namespace means being dependent on your ISP** - what if they renumber or you want to move ISPs? - or going through the bureaucracy of registering your own block. Yes, despite an infinite space you still need to register to avoid fragmentation. That's something I never want to do.
* For example, IPv6 ULAs have a stupid lookup precedence - IPv4 addresses are higher which means ULA can't be used anywhere which is dual stack.
** There are ways around the latter like ULAs (see above) or using a local DNS (more mess if it fails, assuming everything supports this), or NPT - which would work, but again, the community hates anything that reminds of NAT even when it's totally stateless, so this is little documented. I don't think going against the entire community is a worthwhile endeavor. This should also limit your choice of routers since IIRC not everything supports this.
(* Legacy space doesn't even have that)
Not necessarily: as IPv6 becomes more and more mainstream (Google is saying 50% traffic), the need for IPv4 will become less, and so it may be relegated to 'legacy' status and a secondary consideration.