When should I start to panic about IP addresses?
October 18, 2010 5:54 PM Subscribe
How will running out of IPV4 addresses affect the average internet user's Web surfing experience?
I asked my ISP if they were IPV6 ready and they said no. "Blahblahblah is not yet using IPv6 nor are there any immediate plans in place to change to IPv6 at this time."
The subject has been brought up a few times on the blue already:
IANA be a hero, The Future Without IPv6, Uh-oh! We're almost out of IP addresses!
But I'm specifically wondering what the shortage will do to folks. Will it be like the gas shortage in '73? Will we all queue up at godaddy? Should we be hording domains? Should I find a different ISP?
Are we all going to die?
I asked my ISP if they were IPV6 ready and they said no. "Blahblahblah is not yet using IPv6 nor are there any immediate plans in place to change to IPv6 at this time."
The subject has been brought up a few times on the blue already:
IANA be a hero, The Future Without IPv6, Uh-oh! We're almost out of IP addresses!
But I'm specifically wondering what the shortage will do to folks. Will it be like the gas shortage in '73? Will we all queue up at godaddy? Should we be hording domains? Should I find a different ISP?
Are we all going to die?
Best answer: As long as your router supports 6to4 (not all do, but I imagine the list will get longer), then you can access IPv6 sites even if your ISP doesn't support it.
posted by Mwongozi at 5:59 PM on October 18, 2010
posted by Mwongozi at 5:59 PM on October 18, 2010
Best answer: ISPs will turn to NAT, which allows them to put more than one customer behind a given IP address. In the extreme case, they could put all their customers on rfc1918 space and have a single outward-facing IP address that serves everyone. Of course this breaks the end-to-end principle because you cannot accept unsolicited incoming connections behind NAT without either special configuration or special protocol helpers on the NAT machine. But for 99.9% of regular users traffic (web surfing, IMing, etc) you don't need to accept incoming connections.
posted by Rhomboid at 6:24 PM on October 18, 2010 [1 favorite]
posted by Rhomboid at 6:24 PM on October 18, 2010 [1 favorite]
Best answer: Also, domain names have absolutely nothing to do with the issue of IP address allocations, so nobody would queue up at godaddy nor is there anything to horde.
posted by Rhomboid at 6:26 PM on October 18, 2010
posted by Rhomboid at 6:26 PM on October 18, 2010
Best answer: It will probably mostly be a (sometimes painful for techs) transparent transition, but there could be transitory blackholes for some, that is parts of the internet that folks on one ISP can not route to. At some points of growth there is pain, shedding a skin, acne, the web/internet will be going through some teenage years, many parts will be cheerleader(ish) networks and some that weird kid that just doesn't get it, and maybe a few that just get left behind. It also seems like no one really knows just when the transition will happen. Should be interesting.
posted by sammyo at 6:54 PM on October 18, 2010
posted by sammyo at 6:54 PM on October 18, 2010
Best answer: It may (hopefully!) help speed along the demise of IE6.
Many websites can be hosted behind the same IP address using a technique called "virtual hosts". This looks at the browser's request and sends you to the correct site.
With a secure (SSL) site, it can't read this information due to encryption- so many people can't easily use virtual hosts to stick several secure websites behind a single IP address. A technique was invented, Server Name Indicator (SNI), which allows the browser to request the site name before the encryption kicks in.
Unsurprisingly, IE 6 is the last major browser around that doesn't support SNI- so SSL sites and limited IP blocks may help push IE6's festering corpse even closer to the curb.
posted by jenkinsEar at 7:08 PM on October 18, 2010 [1 favorite]
Many websites can be hosted behind the same IP address using a technique called "virtual hosts". This looks at the browser's request and sends you to the correct site.
With a secure (SSL) site, it can't read this information due to encryption- so many people can't easily use virtual hosts to stick several secure websites behind a single IP address. A technique was invented, Server Name Indicator (SNI), which allows the browser to request the site name before the encryption kicks in.
Unsurprisingly, IE 6 is the last major browser around that doesn't support SNI- so SSL sites and limited IP blocks may help push IE6's festering corpse even closer to the curb.
posted by jenkinsEar at 7:08 PM on October 18, 2010 [1 favorite]
IPv6 is no big deal. I got a tunnel and now all my home machines have Internet-routable IPv6 addresses with less setup than it takes to get NAT/DHCP working. If someone wanted to make consumer-level devices that were pre-enabled for IPv6, it would be trivial and every home would have IPv6 in another few years. Same for corporations; it's just a switch to flip somewhere.
Once everyone is using tunnels, ISPs will start offering it natively, and then we won't even need routers at home anymore. (Ok, having a firewall is still nice.)
I don't see this as being a big deal. It will happen when turning on IPv6 makes more money that sitting around and reading AskMe does. Which will be 10+ years, I think.
posted by jrockway at 7:16 PM on October 18, 2010
Once everyone is using tunnels, ISPs will start offering it natively, and then we won't even need routers at home anymore. (Ok, having a firewall is still nice.)
I don't see this as being a big deal. It will happen when turning on IPv6 makes more money that sitting around and reading AskMe does. Which will be 10+ years, I think.
posted by jrockway at 7:16 PM on October 18, 2010
Or peruse the comments on a rather more important mefi-ian's latest blog entry: The internet is full. Go away.
posted by sammyo at 7:17 PM on October 18, 2010
posted by sammyo at 7:17 PM on October 18, 2010
We are all going to die; most of us from natural causes, when we're very old and have lived long and fulfilling lives. IPv4 address space will not figure into that in any way.
posted by togdon at 8:51 PM on October 18, 2010 [1 favorite]
posted by togdon at 8:51 PM on October 18, 2010 [1 favorite]
Response by poster: Wow, all good answers. Thanks folks. I can go back to worrying about global warming now with a clear conscience. (And I really like that it might help rub out IE6. If I could have marked one extra-best that would have been it.)
posted by fartknocker at 8:54 PM on October 18, 2010
posted by fartknocker at 8:54 PM on October 18, 2010
For a more pessimistic take on the subject, this article lists a lot of the difficulties and hurdles that have to be surmounted in order to transition to IPv6. The author thinks that there are going to be, to say the least, some troubles.
posted by Geppp at 9:19 PM on October 18, 2010
posted by Geppp at 9:19 PM on October 18, 2010
NAT and Virtual Hosts for the win.
We were running out of IP addresses real soon now, 15 years ago.
IP6 needed to be implemented real soon now, 15 years ago.
this is a non-issue.
posted by jannw at 1:14 AM on October 19, 2010
We were running out of IP addresses real soon now, 15 years ago.
IP6 needed to be implemented real soon now, 15 years ago.
this is a non-issue.
posted by jannw at 1:14 AM on October 19, 2010
ISPs will turn to NAT, which allows them to put more than one customer behind a given IP address. In the extreme case, they could put all their customers on rfc1918 space and have a single outward-facing IP address that serves everyone.
This actually doesn't work, and NAT isn't a panacea. Every connection from every host behind the translation point requires an outbound port, and there's only 65536 of those per IP. A busy torrent client might be using up to a couple of hundred simultaneous connections, etc. A student house will have a few machines torrenting at a time.
If I were an ISP, I'd never deploy something that prevented customers from having a thousand or so concurrent connections, at an absolute minimum. So that gives you a cap of 64 customers behind each NAT address.
In practice it's a problem for network engineers not end users, but there's no simple solution that doesn't break something for someone.
posted by russm at 2:15 AM on October 19, 2010
This actually doesn't work, and NAT isn't a panacea. Every connection from every host behind the translation point requires an outbound port, and there's only 65536 of those per IP. A busy torrent client might be using up to a couple of hundred simultaneous connections, etc. A student house will have a few machines torrenting at a time.
If I were an ISP, I'd never deploy something that prevented customers from having a thousand or so concurrent connections, at an absolute minimum. So that gives you a cap of 64 customers behind each NAT address.
In practice it's a problem for network engineers not end users, but there's no simple solution that doesn't break something for someone.
posted by russm at 2:15 AM on October 19, 2010
I work at an ISP. We're not ipv6 ready, but we're getting there.
posted by empath at 3:02 AM on October 19, 2010
posted by empath at 3:02 AM on October 19, 2010
If I were an ISP, I'd never deploy something that prevented customers from having a thousand or so concurrent connections
I think you're being overly generous in your estimation of an ISP's mores. Most that I know would consider such a student house to fall under the standard "excessive usage" clause in their AUP and have little regard for their satisfaction, if they didn't try to block the BitTorrent traffic outright. But even that aside, a smart NAT implementation would have a pool of addresses and would assign them dynamically as needed such that there would be no hard limits, so the 64 customers limit is unnecessarily pessimistic. If only 1% of a company's users require more than, say, 500 outbound connections and the other 99% need only a few dozen at most at any given time, then the heavy users are only using a small percentage of the total resources, in much the same way that the ISP can promise 50,000 clients each a 10MBit connection when they only have 1GBit of edge transit serving the region.
posted by Rhomboid at 3:13 AM on October 19, 2010
I think you're being overly generous in your estimation of an ISP's mores. Most that I know would consider such a student house to fall under the standard "excessive usage" clause in their AUP and have little regard for their satisfaction, if they didn't try to block the BitTorrent traffic outright. But even that aside, a smart NAT implementation would have a pool of addresses and would assign them dynamically as needed such that there would be no hard limits, so the 64 customers limit is unnecessarily pessimistic. If only 1% of a company's users require more than, say, 500 outbound connections and the other 99% need only a few dozen at most at any given time, then the heavy users are only using a small percentage of the total resources, in much the same way that the ISP can promise 50,000 clients each a 10MBit connection when they only have 1GBit of edge transit serving the region.
posted by Rhomboid at 3:13 AM on October 19, 2010
According to this Ars Technica article, the numbers are doled out to regional registries, and they are not using them at the same pace. Asia in particular is burning through them. "APNIC and RIPE will have enough addresses to continue business as usual for about another year at current rates, ARIN and LACNIC for around three years, and AfriNIC for nearly seven years. "
posted by smackfu at 6:29 AM on October 19, 2010
posted by smackfu at 6:29 AM on October 19, 2010
You may not always need to buy new hardware. I think ipv6 can be added to most devices with a firmware update, assuming the memory and cpu requirements are met.
posted by empath at 6:39 AM on October 19, 2010
posted by empath at 6:39 AM on October 19, 2010
Same for corporations; it's just a switch to flip somewhere.
That's the funniest thing I've ever heard. You could not be more wrong if you tried.
Anyway, for the average user, I'll go ahead and assume that they won't really notice a difference. ISPs would likely start using private ranges and NAT. Things that required server-to-client connectivity wouldn't exactly work, but there are ways around it.
posted by Threeway Handshake at 6:43 AM on October 19, 2010
That's the funniest thing I've ever heard. You could not be more wrong if you tried.
Anyway, for the average user, I'll go ahead and assume that they won't really notice a difference. ISPs would likely start using private ranges and NAT. Things that required server-to-client connectivity wouldn't exactly work, but there are ways around it.
posted by Threeway Handshake at 6:43 AM on October 19, 2010
This thread is closed to new comments.
posted by damn dirty ape at 5:59 PM on October 18, 2010