Slashdot Mirror
IPv6-only Hosting Won't Make Sense For Years
rawagajah writes "World IPv6 Day this Wednesday will shake out any bugs for websites running on IPv4 and IPv6 in parallel. However, cloud server provider ElasticHosts points out that IPv6-only websites are still a long way off — they only make sense after access is overwhelmingly IPv6 capable. In the meantime, the market in IPv4 space will presumably only grow, benefiting the IPv4 hogs..."
Dual stack hosting does make sense right now, what is slashdot waiting for ?
Right now most residential ISPs don't offer IPv6 period.
We arn't even at the "getting customers to buy into it" phase yet, we are at the "getting it available" point. Which is (and I know this dead horse has been sufficiently beaten already) quite sad considering how much the ipv4 problem has been known about and a solution available.
Even then it probably won't.
Until IPv6 is available to most residential users, it's gonna make more sense to buy ipv4 addresses at high prices than to switch to ipv6 where the huge majority of the internet can't actually get to your site.
The problem here isn't the web hosts, it's the residential ISPs who are _still_ dragging their feet on this.
It's going to make sense when we run out of fucking addresses.
Of course there is the definition of "RUN out" to consider. IP addresses aren't like oil, we don't use them up. When demand exceeds supply then (provided the RIRs don't mess things up too much) they will simply become more expensive causing the least profitable uses to be sacrificed.
I'm betting the first thing to be sacrificed will be public IPs for people on normal home broadband plans (mobile broadband seems to be using private IPs already)
note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register
Comcast is rolling out IPv6 right now in the US. http://www.comcast6.net/
They have had a beta rollout for the past year to work out the issues.
Customers in Denver, CO are currently getting IPv6 to their homes right now.
Hopefully they'll start rolling it out in the San Francisco bay area soon.
What's really sad is that I'm in networking classes right now (not in the Cisco ones yet, that's in a few semesters). And the only mention of IPv6 has been when I asked about it, or asked "how would ___ be done under IPv6?". They aren't even preparing new networking people to work with IPv6. I'm probably going to have to teach myself everything.
OTOH, it doesn't really matter if your non-internet-facing servers are v6 or V4, since they'll only serve local adresses ?
The Cloud - because you don't care if your apps and data are up in the air.
All my websites have been IPv6 ready for many years now and I have never noticed any problems with having them available over both IPv4 and IPv6, but that does not mean there are none. I have read than less than one percent of the users will have IPv6 configured without actually having IPv6 connectivity and I probably loose that traffic. This is what the IPv6 testing day is all about: to see just how much traffic you loose because of badly configured clients. Less than one percent traffic loss may be acceptable to me, but it's not acceptable when you're a huge profitable website. It's pretty obvious that nobody in their right minds will make their high-traffic website available over IPv6 only before 99.5% or so of all users have a IPv6 connection.
9/11: Never forget it was a false-flag operation
"it would give us most likely a good 5 to 6 years to do a nice orderly IPV6 rollout instead of the mess we are in now."
We've had a decade to do a nice orderly IPv6 rollout. The problem is no one will spend the time/money to do it until it is absolutely unavoidable.
If the unused addresses were to be put back into the pool it would give us most likely a good 5 to 6 years to do a nice orderly IPV6 rollout instead of the mess we are in now.
More time isn't what is needed. They've already had lots of time (nearly a decade).
So I say my proposal would buy us the time we need to fix the above problems and make the IPV6 transition a nice slow methodical orderly change over rather than the "Oh shit what are we gonna do?" mess that we have now.
That's the only way it's gonna happen. Like many other problems (pollution or fossil fuel) that cost a lot of money to fix just to get back to nominal, it's not gonna be dealt with until stuff starts actually breaking.
and how big are your routing tables?
The Grey Goo disaster happened 3 billion years ago. This rock is covered in self replicating machines!
And there's the same problem with nobody buying the IPv4 2.0 kit because nobody uses IPv4 2.0
And it doesn't have the built in cool stuff which nobody has worked out that they want yet (Mobile IPv6 + ubiquitous IPSec means no more half-arsed VPN software - hooray!)
And who wants the knowledge base to stay the same? If there's a better way to do things, lets do it!
Do you really think there isn't a "cluster fucking 35 different versions of shit and different libraries" on your windows box?
If you really believe that, I would like to invite you to check out %windir%\WinSxS; it is part of a mechanism designed to resolve traditional Windows DLL hell but can become VERY bloated over time It's where system libraries are actually stored and then are linked to from other directories. Due to the past DLL hell, it is rare that anything ever gets deleted from WinSxS in order to prevent DLL hell by inadvertantly deleting a library that might be marked by the registry as unused, but is actually relied upon by a seldom-used app. So, what happens is as you install and upgrade your various applications, system drivers, and whatnot, a ton of files often get written to in WinSxS when installers don't check for dependencies - how many times have applications forced installs of components you know are already in place? Why does this happen? Because all too many release engineers don't understand system administration, how the OS works at the low level, so they don't know how to check for preexisting components. Why is this? Because hiring managers are all too focused on specific tool (Rational Clearcase and Clearquest, Installshield, Visual Studio, Ant, Eclipse, or a specific language, etc) and not on what really matters, i.e., system administration, coordinate development and QA, manage the build platform and a build a clean net, etc. Too much emphasis is based on knowing a specific application, rather than the process and ability to learn a tool quickly. Individual tools are relatively easy to learn very quickly; system administration and basic scripting skills are relatively difficult to pick up quickly. I never focused on learning all the tools out there; I learned the individual tools as I needed to, so my installers were always rock-solid because I knew the requirements for the underlying system, and my installers didn't force unnecessary component updates which bloat a system.
So, your Windows vs. Linux argument is kind of moot; you may not realize it, but even though you might not see libfoo.so.0.2.1, libfoo,so.0.2.1 and libfoo.so.0.4.1 (and a symlink from libfoo.so.0.4.0 to libfoo.so.0.4.1 since it's compatible and the install creator decided to save you space but not break your system in the process) in /usr/lib on Windows, but if you have installed and over time upgraded various applications you easily have 5 to 10 different copies of various libraries - often identical versions, cluttering up WinSxS.
Check these out:
http://www.ghacks.net/2010/07/24/the-winsxs-folder-explained/
http://blogs.technet.com/b/askcore/archive/2008/09/17/what-is-the-winsxs-directory-in-windows-2008-and-windows-vista-and-why-is-it-so-large.aspx
Unix-based systems are easy to clean, maintain, and if you do break /usr/lib, very easy to fix in comparison to Windows. Now tell me - after reading those articles, if you have the Unix experience you claim to have, after learning how Windows deals with various library versions, which system is better and more logical? Don't get me wrong; Microsoft has done a fantastic job making Windows a hell of a lot more stable than it used to be, but this "fix" is still a major hack which doesn't fix the root problem: shitty release engineers not developing and adequately testing installers until they're rock solid.To work around install developer incompetence, we have come to a point where WinSxS may contain gigabytes' worth of old cruft that is no longer used on a Windows box.
The Christian Right is Neither (Christian nor right). See: Matthew 23, Matthew 25, Ezekiel 16:48-50
There are after all 4 billion IPv4 addresses
It would intially seem that way. However when we look at the 256 /8 blocks we see that many of them are not normal IPv4 addreses.
1 block is assigned to local identification /8)
1 block is assigned to private use (there are also private use blocks elsewhere but only one is a full
1 block is reserved for local loopback
16 blocks are allocated to multicast
16 blocks are practically unusable because they were never defined as either unicast or multicast and IIRC windows won't accept them as addresses
So there are only arround 3.7 billion "regular" addresses rather than the approximately 4.3 billion that one would nievely expect. Further conventional subnetting wastes quite a few addresses too, you waste one for network, one for broadcast, one for gateway and however many are needed to make the number of addresses up to a power of two. So i'd imagine the number of usable addresses is more like 3 billion.
And the cynic in me tells me it's not going to be our home broadband plans
What I expect will happen in the west is that public addresses will gradually (it will vary a lot by the particular ISP's growth rate and address situation at the time of exhaustion) become an extra cost option. If an ISP charges a couple of bucks extra a month for one then they are likely to free up a lot of addresses without pissing off the geeks too much.
One thing that is not clear at this point is whether it will be possible for ISPs to sell addresses across RIR boundries or if sales will be restricted to one RIR's area.
note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register
"it would give us most likely a good 5 to 6 years to do a nice orderly IPV6 rollout instead of the mess we are in now."
We've had a decade to do a nice orderly IPv6 rollout. The problem is no one will spend the time/money to do it until it is absolutely unavoidable.
This.It wouldn't make a difference, as it would just mean everyone would continue doing nothing, and legitimate users would just pay more.
/27 for free on my home network and I enjoy not having to use NAT and I am using the addresses (well more than 16 of them). Now why should I have to pay an extra $30 for my net connection because the rest of the Internet providers haven't performed due diligence with this issue (and since my ISP has also been IPv6 ready since 2002 they are obviously doing their job properly)
My ISP gives me a
Or gradualy the routing tables will get out of the reach of the routers at some places, and IPv4 will completely stop working.
There are many problems with auctioning IP addresses.
Rethinking email
I use IPv6 (alas via a tunnel here). IPv6 gets rid of DHCP, which is nice, but also lets me ssh directly to any of my machines here. So I can ssh to my wife's machine to fix her machine (often that the old box has too much flash running), and git pull those changes I forgot to push from my home computer. Very convenient.
Religion is regarded by the common people as true, by the wise as false, and by rulers as useful.
Not on every machines. On the router, this is enough. I just bought a D-Link DIR-825, and re-flashed it with OpenWRT. Using a 2GB key as root filesystem, it's working very well, and it has all the facilities needed (eg: ip6tables) to block any bad incoming connection. It also has all the software you need to get yourself connected to the ipv6 network, even if your ISP doesn't provide it: you can connect to an ipv6 capable host using OpenVPN, or setup an ipv6 in ipv4 tunnel if you like (for example, from SixxS, or others). I know it may sound like top-advanced, but frankly, it's not that hard to setup. Sure, many people wont do it. But for us, IT people reading slashdot, it's really possible and easy.
As for the firewall, people are badly mistaking NAT with a firewall. Don't do that, it's dangerous. Soon, you'll get your laptop connected to the net directly (maybe using 3G?), and you'll regret your wrong thinking! The fact that masquerading acts as a firewall is pure luck.