Huston: Another 10 years later
Perhaps this increased resistance to change is because as the size of the network increases, its inertial mass also increases. We used to quote Metcalf’s Law to each other, reciting the mantra that the value of a network increases in proportion to the square of the number of users. A related observation appears to be that a network’s inherent resistance to change, or inertial mass, is also directly related to the square of the number of users as well."
Posted Jun 29, 2018 4:21 UTC (Fri)
by zaitcev (guest, #761)
[Link]
Posted Jun 29, 2018 7:39 UTC (Fri)
by jem (subscriber, #24231)
[Link] (12 responses)
IPv6 is here already, and is not going away. For example, according to Google, 37% of US and German users access Google using IPv6, and 50% of all users in the US and India access Facebook using IPv6. Btw, India has the largest population of IPv6 users in the world.
On the other hand, there are huge differences in adoption from country to country. Countries with large populations in Asia are seriously lagging behind the rest of the world, and the whole continent of Africa (except for Zimbabwe) is below 1%. There are big differences within countries, too. Some ISPs have embraced IPv6, others apparently couldn't care less.
Back in the years when the Internet was growing fast, there was much talk about how big growth the net could sustain without fragmenting. I fear the Internet is again at risk to break into two: the IPv6-enabled part where IPv6 is gradually taken for granted with IPv4 being the legacy protocol being phased out, and the part where work on the transition has barely begun. The world has already run out of IPv4 addresses, and I predict there will be a point in time when some web sites start thinking it's not worth the cost to serve the whole world, and will go IPv6 only.
Posted Jun 29, 2018 16:31 UTC (Fri)
by excors (subscriber, #95769)
[Link] (11 responses)
It being (partly) here doesn't mean that we need it. We thought we needed it to prevent the IPv4 address exhaustion apocalypse, but the apocalypse happened a few years ago and nobody really noticed. The internet has continued working okay on primarily IPv4. That seems to invalidate the main motivation behind IPv6, and the migration cost is enormously higher than anyone predicted, and it will continue to have a high cost in the future (for ongoing migration work, plus either the cost of dual-stack support or the harm of network fragmentation), so maybe it's time to forget about the sunk cost and abandon it as a failed experiment.
Posted Jun 29, 2018 18:22 UTC (Fri)
by pizza (subscriber, #46)
[Link]
> That seems to invalidate the main motivation behind IPv6, and the migration cost is enormously higher than anyone predicted, and it will continue to have a high cost in the future
Folks sitting behind three layers of NAT might dispute your "working okay" assessment.
...And you seem to think that "continued working okay" did not come with an ever-increasing price tag of its own.
> so maybe it's time to forget about the sunk cost and abandon it as a failed experiment.
It's not just the sunk cost that keeps folks plugging along with IPv6. It's that there's no credible alternative, and even if there was, it's going to be even more expensive (in both time and money) to get it deployed to the point of general usefulness as we'd be starting from scratch. Again.
Posted Jun 29, 2018 18:34 UTC (Fri)
by farnz (subscriber, #17727)
[Link] (4 responses)
The question is whether the engineering cost involved in sustaining the systems that extend IPv4's lifetime are greater than or less than the cost in going to IPv6.
This is where things like Comcast moving to dual stack and Reliance Jio offering IPv6 + 464XLAT only are interesting - are they seeing savings from reduced IPv4 infrastructure that outweigh the costs of their IPv6 infrastructure? With Google (thus YouTube), Netflix, Facebook etc being IPv6-enabled, is it the case that you get a majority of traffic on IPv6, and the NAT (NAT44 or NAT64) needed to sustain IPv4 is becoming less and less loaded over time, and thus less critical to "important" use cases?
Note that as long as ISPs can get enough IPv4 to assign one per customer, we're not in the full-on apocalypse phase; the apocalypse does not start until ISPs have to tell customers "you're behind CGNAT for IPv4 service" and thus have to make the IPv4 NAT actually reliable to the extent users expect. My experience is that this is damnably expensive - for IPv6, the state sharing needed to ensure that all firewalls in a network will let at least the user's desired traffic reach their handset is much smaller than the state sharing needed to ensure that all CGNATs in a network will sustain all translated sessions for a given user.
Posted Jun 30, 2018 2:05 UTC (Sat)
by Cyberax (✭ supporter ✭, #52523)
[Link] (3 responses)
Posted Jul 3, 2018 8:02 UTC (Tue)
by tialaramex (subscriber, #21167)
[Link] (2 responses)
Posted Jul 3, 2018 17:45 UTC (Tue)
by Cyberax (✭ supporter ✭, #52523)
[Link] (1 responses)
What is more problematic, are the network organization headaches. Voice-over-LTE requires IP connectivity, since it's based on SIP. Contrary to many beliefs, LTE can work just fine with a purely IPv4 network but you do start having problems at scale. Like getting enough IPs for even your network equipment to do end-to-end connectivity between edge SIP servers. And while SIP in theory can cope with various types of NATs, its implementation inside phone baseband chips is sometimes lacking robustness.
Posted Jul 4, 2018 9:15 UTC (Wed)
by farnz (subscriber, #17727)
[Link]
From what friends at mobile ISPs tell me, the cost of CGNAT is not in the equipment; it's in the IP reputation management costs, of which the big elephant in the room is Google and its family of sites (Blogger, YouTube etc). The problem is that these sites don't distinguish a large CGNAT with a lot of users from a single user behind an IP with a bot scraping the sites, and start putting up interstitials talking about the botnet issue and asking you to confirm that you're a human. Those interstitials worry humans, who then phone in for support, believing that their phone has been hacked, so you have to do something to keep the IP reputation high enough to avoid those interstitials.
IPv6 already mitigates that - Google et al can see that there's a wide distribution of IPv6 addresses in use, and don't trigger as quickly. Plus, their process appears to be "interstitials for a single /128 first, then escalate if the bot-like behaviour does not stop".
Posted Jun 29, 2018 19:14 UTC (Fri)
by Cyberax (✭ supporter ✭, #52523)
[Link]
Meanwhile, IPv6 costs are dropping. All the modern operator-level gear supports, the problem is in the leaf ISPs and consumer devices.
Posted Jun 29, 2018 23:54 UTC (Fri)
by Wol (subscriber, #4433)
[Link] (3 responses)
At which point the shift to IPv6 really started taking off. I think my phone runs v6.
Thing is, the whole idea behind the Internet is that your computer should be PART OF the internet. If you're behind NAT then you're just a CLIENT OF the internet. Which means for the Internet to function as designed then IPv6 is now a necessity.
As commented elsewhere, as long as people could soldier along without v6, they did because the cost of switching (both money and time and experience) was quite high. As private networks started switching - either home enthusiasts or businesses struggling to cope as the v4 shortage started to bite, the cost of switching fell.
And v6 is now really on the up, because while the cost of new v6 deployments fell, the cost of new v4 deployments rose. And address exhaustion was the trigger that made the cost of v4 hike to where it is now probably higher than v6.
Think of it like digital TV, especially in Europe. Americans switched quickly, because 720p was much better than analog NTSC. That drove the price down, and then Europeans started switching - they were a lot slower because 720 isn't that much better than analog PAL. Then the providers began to shut down the analog transmitters because the number of analog TVs out there began to drop.
These changes take time - for ages nothing appears to be happening, then all of a sudden it seems to switch "just like that". The cost-benefit curve is now firmly in v6's favour.
Cheers,
Posted Jul 5, 2018 16:22 UTC (Thu)
by BenHutchings (subscriber, #37955)
[Link] (2 responses)
Posted Jul 7, 2018 15:31 UTC (Sat)
by Wol (subscriber, #4433)
[Link] (1 responses)
Okay, I'm not that interested in TV, so you are probably right. But the point still holds that if i've got a fully functional 625-line analog system, I'm unlikely to want to spend a lot of money upgrading to 625 or 720 digital.
Cheers,
Posted Jul 9, 2018 0:39 UTC (Mon)
by anselm (subscriber, #2796)
[Link]
Which is why people had to be dragged kicking and screaming into the age of digital TV. Here in Germany, there was a flag day where the analog TV transmitters were simply switched off; if you wanted to continue watching TV, you had to buy a new digital-enabled TV or at least a set-top box, or all you'd get to see would be channel noise.
Posted Jun 29, 2018 17:17 UTC (Fri)
by excors (subscriber, #95769)
[Link] (3 responses)
I'm not sure about that observation, for any normal meaning of "users". Facebook is very valuable because it has a lot of users, but Facebook can make changes to their platform pretty easily. Google has a lot of users, and invents and deploys plenty of new internet protocols - they can push things like HTTP/2 and QUIC to a billion Chrome users overnight. Resistance to change doesn't depend on number of users, it depends on the number of independent groups that have to put effort into making that change happen. Proprietary systems can innovate relatively easily since one group can take unilateral action, while widely distributed systems with no central ownership will inevitably stagnate. Unfortunately the IP-level internet is an example of the latter case, but that's a problem of design rather than a problem of size.
Posted Jun 29, 2018 23:40 UTC (Fri)
by Wol (subscriber, #4433)
[Link] (2 responses)
Which means that innovation has great difficulty taking place at the IP level, because the hardware actively obstructs it ...
Cheers,
Posted Jul 1, 2018 5:12 UTC (Sun)
by willy (subscriber, #9762)
[Link] (1 responses)
Posted Jul 1, 2018 13:56 UTC (Sun)
by corbet (editor, #1)
[Link]
Huston: Another 10 years later
Huston: Another 10 years later
Huston: Another 10 years later
>
> IPv6 is here already, and is not going away.
Huston: Another 10 years later
Huston: Another 10 years later
Huston: Another 10 years later
Huston: Another 10 years later
Huston: Another 10 years later
Huston: Another 10 years later
Huston: Another 10 years later
Huston: Another 10 years later
Wol
Huston: Another 10 years later
Huston: Another 10 years later
Wol
Huston: Another 10 years later
Huston: Another 10 years later
Huston: Another 10 years later
Wol
Huston: Another 10 years later
Coverage of the LCA talk, complete with link to the video.
QUIC