For server Focus on Linux?

While I'm sure a few people / companies have implemented IPv6 for the office LAN it seems somewhat pointless to me. Any local LAN network is going to have it's own local DHCP serving from a local IPv4 pool with a router to a possible IPv6 WAN address. For most that's probably still going to be an IPv4 address with it getting translated to IPv6 further up the line if needed. Many consumer devices out there don't even support IPv6 still, like any iOS device not upgraded to iOS 6.

Also, while XP might not support IPv6, no one should be directly connecting an XP system to the Internet with out a router in between... though I'm sure a ton of people still do.

I can personally vouch for this. I am an admin at a colo/cloud company. We operate 4 datacenters. IPv6 only addresses are completely unacceptable to our customers. The reason why is that if there are locations that do not support it and many end points don't, They can't get to their remote offices. Further adding to the problem is that many routers currently deployed at SOHO locations do not support the protocol, even though all current OS's and many major ISP's do. Long story short were going to be running dual stack for a very long time, as in decades! The price for a class c is just going to keep climbing.

While RFC 6144 allows for NAT between v4 and v6 addresses the entire point of IPv6 is to get rid of NAT entirely per RFC 6296. Ideally each host should have one or more publicly routable addresses configured directly on its NIC with a stateful firewall providing security either on the host itself or upstream at the network level.

Also some "LANs" have more hosts than can fit in the RFC 1918 space. This makes deploying IPv6 on the LAN a priority as the global address pool runs out.

iOS has supported IPv6 since version 4 and Android has supported it since at least 2.2. Most higher end consumer routers support it (although people don't upgrade these very frequently, which is an issue) and pretty much everything in the enterprise market supports it. The real issue is that residential ISPs have no control over customer equipment past the demark point and still haven't figured out how they want to handle prefix delegation. This is actually one of the reasons why ISPs like AT&T U-verse are pushing all in one NAT gateway boxes so hard, it lets them control your home router which makes deploying IPv6 easier for them.

Dual stack is certainly the way to go though and I recommend it to, and deploy it for, my customers all the time. Deploying an IPv6 only environment outside of a lab is just dumb at this point in the global IPv6 deployment process.

Once this price exceeds the price of deploying IPv6 for a small to medium sized business we will hit a tipping point that will start an avalanche of IPv6 deployments.

The usual local range of 192.168.###.### has allotments for 65534 unique IPs, there aren't a lot of businesses that will ever use that many. Granted most system stick to just the same subnetwork (limiting the number of unique IPs to 254), which is something I've seen causes issues before. But really if you've got a local area network big enough to need more than 65536 unique IPs, you're probably running some kind of server farm. In that case you're likely already using IPv6 or using multiple layers of NATs (seen this at one of Cisco's old test labs, so many routers to routers to routers to routers...) or completely segregating your main server network and just using the full range of IPv4 with a hand full of systems exposed to the rest of the office network handling requests.

On a funny note, some console devkits I've used make assumptions about the subnet mask being 255.255.255.0, which can cause some headaches when the DHCP eventually decides to put the devkit and your workstation on different subnets.


But yes, IPv6 is the eventual future people should be working towards, but I doubt we'll see it go away for a long, long time.

I would beg to differ. A single /16 really isn't that big when you start doing isolated FTP, load balancing, SSL, and HA left and right. And you would be amazed at the things that need an IP that are not PCs, printers, or smartphones/tablets. Did you know that digital cable and satellite TV boxes all have IP addresses?

I have seen plenty of corporate networks that use the entire 10.0.0.0/8, 192.168.0.0/16, and 172.16.0.0/12 spaces and still need more IPs. Heck I have seen just a printer vlan use an entire /16, for just printers. When this happens and you need to be able to talk to the internet the only solution is to go to your local Regional Internet Registry and request some IP space. If there isn't any available then you are kinda screwed. You can mess around with clever policy NAT trickery but that really doesn't scale very well and its a nightmare to support and maintain.

Not necessarily, some organizations have an awful lot of workstations and end user terminals. How many cash registers do you think McDonalds has globally? Just because its a "LAN" doesn't mean that it doesn't span several continents.

NAT is really not an option with Active Directory, passive FTPS, VoIP, and tons of other protocols that are common on a large LAN. Also since very large chunks of any LAN tend to need internet access you generally can't totally isolate yourself and play in your own little walled off universe either.

Deploying IPv6 across the entire network is not always feasible within the time and budget constraints set forth by business reality. IPv6 enabled printers are still awfully rare and replacing tens of thousands of printers is quite expensive, time consuming, and very hard to justify to management if they all still print just fine.

That is truly silly. I admit that /24 ranges are certainly the norm in the SOHO world but not all companies that develop console games are small.