Stories
Slash Boxes
Comments

SoylentNews is people

posted by janrinok on Saturday July 06, @06:42PM   Printer-friendly

Back in 2018 the Internet Engineering Task Force (IETF) announced that the IPv6 protocol had become a full Internet standard:

With IPv6 adoption accelerating over the past 6 years, from being a negligible fraction of the Internet (<1%) to recently topping 25% [Ed., now over 42%], moving IPv6 a full Internet Standard could not have come at a better time.

The Internet Standard designation represents the highest level of technical maturity and usefulness in the IETF standardization process. As the relative numbering of the RFC (RFC 8200) and STD (STD86) suggests, there are many protocols that make their way through the IETF standards process to be published as RFCs, but are not Internet Standards. The Internet Standard designation means those implementing and deploying a protocol can be assured it has undergone even more technical review by the IETF community than the typical RFC, and has benefitted from experience gained through running code and real-world experience. This is definitely true in the case of IPv6.

[...] Moving these IPv6-related specifications to full Internet Standards matches the increasing level of IPv6 use around the Internet. The IETF community has steadily worked to ensure that the Internet is ready for the time when IPv6 is the dominant Internet Protocol. Work in a variety of IPv6-related IETF working groups, such as 6man and 6ops, continues, striving to make the Internet work better.

On 02 July the IETF Executive Director announced that they have given up on IPv6 as being too much effort for their own services starting with email:

3. IPv6 for mail
As others have explained, we have chosen to switch, at this stage, to a large commercial mail sender with extensive reputation management rather than continue to send directly and as a consequence that will be IPv4 only. I don't plan to reiterate the multiple trade-offs considered in that decision, but I do want to stress that this was not a simple decision. I say "at this stage" because there are still discussions about whether or not this is the best long term strategy for mail delivery.

At a principled level, I agree that if the community says "we must have IPv6 for mail" then the LLC needs to deliver that, but at a practical level, given the cost and effort required, I would want that conveyed in a more formal way than a discussion on this list and us given a year plus to deliver it. However, and this is major however, piecemeal decisions like that are only going to make things much harder and it would be much better to have a broader decision about IPv6 in IETF services (more on that below).

For now at least then, we are going to continue with the plan to move to Amazon SES for mail sending. Once that is bedded down, that will be reviewed, but that will be several months away and the outcome may be to stick with it, unless there has been a community decision that changes that.

4. IPv6 for all services (or not)
If the community wants to develop guidance on the use of IPv6 for IETF services then that would be helpful. More generally, it would be so much better all round, if the implicit expectations that people have about IETF services, were properly surfaced, discussed, agreed and recorded. If that were done, then we would be very happy to include those in any RFP or service assessment.

Will people care if the organization who, for very many years, has been strongly advocating for everyone to switch to IPv6 has now given up on it? At a superficial level it doesn't look great if that decision was effectively made by AWS.


Original Submission

This discussion was created by janrinok (52) for logged-in users only, but now has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
(1)
  • (Score: 4, Informative) by stormreaver on Saturday July 06, @09:01PM (1 child)

    by stormreaver (5101) on Saturday July 06, @09:01PM (#1363328)

    The headline is very misleading. The article doesn't say that transitioning to IPv6 is difficult and expensive because of IPv6. It says the IETF is facing a bureaucratic situation that is generating difficulty and expense. To satisfy the bureaucracy, they are moving their email servers to Amazon for now.

    • (Score: 2) by loonycyborg on Saturday July 06, @09:29PM

      by loonycyborg (6905) on Saturday July 06, @09:29PM (#1363330)

      Still there's lot of ipv6 enabled mail providers. Probably they're just going with Amazon stuff because they happen to have more experience with it. But dogfooding is important too, so finding something with ipv6 is worth it I think.

  • (Score: 3, Interesting) by Ox0000 on Saturday July 06, @09:46PM (14 children)

    by Ox0000 (5111) on Saturday July 06, @09:46PM (#1363333)

    Just for clarity: this is a genuine question borne out of my ignorance on IPv6.

    I understand some of the things that IPv6 enables, which seems to boil down to mostly to offering a larger address space and enabling 'interesting' QoS scenarios (I understand there are others others, and that those are less important - again, I'm rather uninformed on IPv6). But my real question is: what actually makes IPv6 so hard (full stop as well as to migrate to)? Can someone give me a primer so that I (finally) start understanding what the big deal with it is?

    • (Score: 5, Interesting) by RS3 on Saturday July 06, @10:12PM

      by RS3 (6367) on Saturday July 06, @10:12PM (#1363335)

      I don't have a definitive nor complete answer, but some of the problem is simply legacy software, drivers, and firmware such as in gateway/routers, which don't do IPv6.

      People love to say "upgrade". Easy to say, can be costly or very costly to do. Case in point: I admin a (very) small hosting operation. Owner has no motivation to upgrade anything. He makes no $ from it- he keeps it running for convenience because he (we) hosts some of his own websites there, and some for family / friends' sites, and some paying clients.

      That said, ISP has sent several upgrade gateways for free. I tried all but the last one. None of the several that I tried will let me static map Internet addresses and ports to internal non-routeable (192.168...) ones the way the older one will.

      Bottom line: it works as is, and somewhere in the machinations of the Internet something translates IPv6 to IPv4 automagically.

      Windows has software called "Teredo" that does among other things IPv6 to IPv4 translation.

      My take on it: the IPv4 to IPv6 translation mechanisms work well, there's little incentive to fully abandon IPv4.

    • (Score: 5, Insightful) by gnuman on Saturday July 06, @10:15PM (4 children)

      by gnuman (5013) on Saturday July 06, @10:15PM (#1363337)

      Nothing makes it hard. But people are lazy and will not move, like a donkey.

      * How long was `telnet` the standard way of login in? I remember at univerity, ssh was not even installed. Then admins complained that "hackers" can easily break into this Linux so they better stick with Solaris.
      * How long was https:// too heavy for websites? until ISPs starting inserting ads into 3rd party sites, then suddenly it became not heavy
      * For how long is DNSSEC deemed as "not important"?

      My ISP will not deploy IPv6. Why? Because they have enough IPv4 so why would they bother? Another, more successful ISP around here, doesn't even offer native IPv4 anymore because they don't have the addresses to give out. Carrier grade NAT only for IPv4s, otherwise, IPv6 native.

      Finally, it pays to have IPv4 around. ISPs are charging $$$ for scarcity (ie. the server side). This actually is an incentive to keep IPv4 around longer!

      Does this answer your question?

      • (Score: 0) by Anonymous Coward on Sunday July 07, @06:48PM (3 children)

        by Anonymous Coward on Sunday July 07, @06:48PM (#1363381)

        My ISP will not deploy IPv6. Why? Because they have enough IPv4 so why would they bother? Another, more successful ISP around here, doesn't even offer native IPv4 anymore because they don't have the addresses to give out. Carrier grade NAT only for IPv4s, otherwise, IPv6 native.

        If you are providing a service of any kind on the internet, you basically have no choice but to make it accessible via IPv4. It remains extremely inexpensive to do so, so literally every service of any value on the internet is available on the IPv4 internet.

        There is no actual value today in having a connection to the IPv6 internet. So there is no reason for an ISP to provide it, except as a mechanism to access the IPv4 internet and perhaps to appease a handful of computer nerds (which might work to generate some business, but otherwise is probably not something to spend a lot of resources on).

        ISP customers as a whole will demand IPv6 access only when there are significant internet services which are only available via IPv6. If that ever happens, then IPv6 will suddenly become very important. But until then, it is actively counterproductive: everything works better on IPv4, and having a fully dual stacked connection to the internet can certainly introduce problems but does not solve anything.

        • (Score: 3, Informative) by janrinok on Sunday July 07, @07:28PM (2 children)

          by janrinok (52) Subscriber Badge on Sunday July 07, @07:28PM (#1363388) Journal

          I am guessing that you are in the USA. Here in Europe most things are IPv6. The websites that I access, the email servers that I access in UK, France and elsewhere. My entire home network is IPv6. For me, it changes when I want to access this site. At some point the routing usually goes to Paris and then it is converted to IPv4 when it goes outside Europe (I think). If I query 'WhatsMyIP" it gives me an IPv6 response.

          There is probably a technical name for this conversion at a gateway of some sort set up by Orange.fr, but for me I don't need IPv4. If I type 'ip a' at the command line I get lots of IPv6 information, but nothing that is IPv4.

          Internally our whole SN site is IPv6 enabled and has been that way for at least 5 years, although our servers still seem to have IPv4 addresses too. For example, a France-wide supermarket chain (Carrefour.fr) gives me the following.

          PING carrefour.fr(2606:4700:4400::ac40:9b27 (2606:4700:4400::ac40:9b27)) 56 data bytes
          64 bytes from 2606:4700:4400::ac40:9b27 (2606:4700:4400::ac40:9b27): icmp_seq=1 ttl=56 time=14.7 ms
          64 bytes from 2606:4700:4400::ac40:9b27 (2606:4700:4400::ac40:9b27): icmp_seq=2 ttl=56 time=14.3 ms
          64 bytes from 2606:4700:4400::ac40:9b27 (2606:4700:4400::ac40:9b27): icmp_seq=3 ttl=56 time=15.0 ms
          64 bytes from 2606:4700:4400::ac40:9b27 (2606:4700:4400::ac40:9b27): icmp_seq=4 ttl=56 time=14.4 ms

          I am still on ADSL.... Fibre is some 3-5 years away from rural France although every significant town and city has it. Just not us country yokel types.

          --
          I am not interested in knowing who people are or where they live. My interest starts and stops at our servers.
          • (Score: 0) by Anonymous Coward on Monday July 08, @10:19PM

            by Anonymous Coward on Monday July 08, @10:19PM (#1363494)

            I am guessing that you are in the USA. Here in Europe most things are IPv6. The websites that I access, the email servers that I access in UK, France and elsewhere. My entire home network is IPv6. For me, it changes when I want to access this site. At some point the routing usually goes to Paris and then it is converted to IPv4 when it goes outside Europe (I think). If I query 'WhatsMyIP" it gives me an IPv6 response.

            I did not mean to suggest that there is nothing at all on the IPv6 internet. Obviously there are lots of sites out there that can be accesed with IPv6. The point is that nobody actually cares about IPv6 connectivity, because every service anyone cares about is also on the IPv4 internet, and only computer networking geeks are going to care one whit whether that connection was done over IPv6 or IPv4.

            Internally our whole SN site is IPv6 enabled and has been that way for at least 5 years, although our servers still seem to have IPv4 addresses too.

            Yes, exactly.

            If you deleted the AAAA records for soylentnews.org today and then (later) turned off IPv6 support on the servers, you would not cause any noticeable problem for any real visitor to this website.

            The reverse is not true. If you deleted the A records you will suddenly have a lot of users with connection problems.

          • (Score: 1) by pTamok on Tuesday July 09, @04:56PM

            by pTamok (3042) on Tuesday July 09, @04:56PM (#1363560)

            janrinok: nope.

            I'm forced to use two different ISPs in two different European countries (due to the blocks of flats having a management group that choose the ISP that serves the block).

            In one country, the ISP stopped offering IPv6 late last year, with no timeline for reinstating it.

            In the other country, the ISP renumbered IPv6 without telling me, cutting me off from remote access to my router. They, by policy, provide no help in connecting up my own router again, saying they support only their own router (which is managed/backdoored by TR-069) - they also look to have gone down the path of offering only a /64 to 'customers'.

            I'm spitting feathers.

    • (Score: 2) by loonycyborg on Saturday July 06, @10:15PM

      by loonycyborg (6905) on Saturday July 06, @10:15PM (#1363338)

      It's not very hard, but might involve lot of changes and lot of testing for large scale organizations. All client systems most people use already support ipv6 natively. Providers show more inertia, both because it's lot small things to test and fix and because the motivating issue of ipv4 address space exhaustion isn't as pressing yet.

    • (Score: 2, Insightful) by Anonymous Coward on Saturday July 06, @11:19PM

      by Anonymous Coward on Saturday July 06, @11:19PM (#1363346)

      what actually makes IPv6 so hard (full stop as well as to migrate to)?

      I worked at an internet provider over 3 decades ago. It was my second job out of high school.

      I still remember the IP of their main router and the IPs of all the routers that the POPs they had at the time.

      I still remember the IP of the mail server, DHCP server, DNS servers, and the core switch.

      3 decades later and their main router and all the POPs still use the same IP address.

      Now if I could remember my 37-hex-character IPv6 address, my netmask, the address of my router, and the address of my DNS server off the top of my head, I can get started on the switch.

    • (Score: 4, Informative) by sigterm on Sunday July 07, @02:00AM

      by sigterm (849) on Sunday July 07, @02:00AM (#1363353)

      It's not at all hard to "migrate" to IPv6, mostly because you're not actually migrating anything, you're just running IPv4 and IPv6 in parallel (dual stack).

      The procedure looks like this:

      • Make sure your OSes support IPv6: They all do since around 2010.
      • Make sure your routers and firewalls all support IPv4: You may have to replace ancient equipment.
      • Obtain a suitable number of IPv6 addresses: Not an issue if you're a LIR/ISP; if not, decide whether you're happy with using your provider's address space (most are) or if you need provider-independent addresses, which involves registering an AS number and running BGP on your border routers, just like with IPv4 (if you really need provider-independent addresses you're probably running BGP already).
      • Assign addresses as needed: It's the same procedure as with IPv4, except gateways are detected automatically.
      • Create the requisite AAAA DNS records to make your services available via IPv6.

      Congratulations, you're now IPv6 compliant. I've done this more times than I can count, and it really is a very straightforward procedure.

      As for advantages compared with IPv4 they are numerous and fairly major, the most obvious being that IPv6 offers a vastly expanded address space, which allows both massive address ranges for every endpoint as well as strictly hierarchical routing, which greatly simplifies backbone routing and reduces the size of the routing tables by at least two orders of magnitude compared with IPv4.

      The latter is achieved by IANA allocating almost comically large address ranges to the respective RIRs, which in turn can give out huge address ranges to the LIRs, which then hands out at least a /64 to every subscriber, with /48's typically being available for free upon request. So rather than having to NAT your devices behind a few IPv4 addresses (or even one), you always get a network 2^64 addresses to play with. You can subnet this if you like but if you can't be bothered, just ask for a /48 and you'll have 65536 such /64 networks at your disposal.

      Another major improvement is that all IPv6-compliant routers support multicasting right out of the box. If ISPs are willing to play ball, this has the potential to revolutionize streaming and software distribution.

    • (Score: 2) by bzipitidoo on Sunday July 07, @02:37AM

      by bzipitidoo (4388) on Sunday July 07, @02:37AM (#1363355) Journal

      IPv6 is more efficient. Although the addresses are larger, IPv4 has some cruft that was dropped in IPv6. I forget exactly what fields IPv4 requires that designers decided wasn't actually necessary or useful.

      IPv6 supports much, much larger packets. IPv4 had a maximum packet size of 64k, though often packets were only 1.5k. That was fine in the days when networks were much slower and less reliable, but today, such tiny packets are overkill that clutters up the bandwidth with overhead. IPv6 supports packets up to 4G in size.

      IPv6 is also a lot better for multicasting. Anyone can receive a braodcast just by getting an address in the /64 subnet the broadcaster is using.

    • (Score: 3, Insightful) by driverless on Sunday July 07, @02:41AM

      by driverless (4770) on Sunday July 07, @02:41AM (#1363356)

      I think the best way to handle it for most users is to do nothing and let it take care of itself. Over time things that handle IPv6 will silently switch to IPv6 and things that don't will continue with IPv4. I've noticed that with my IT gear, every now and then I'll see that something is using an IPv6 address when six months earlier it was IPv4. Nice, it's upgraded itself in the background, and things continue as before.

    • (Score: 3, Interesting) by bryan on Sunday July 07, @08:09PM (2 children)

      by bryan (29) <bryan@pipedot.org> on Sunday July 07, @08:09PM (#1363395) Homepage Journal

      The problem is similar to when CPUs were switching to 64 bit. The largest power in the industry (Intel) decided to throw the existing (x86) architecture away and make a new architecture (Itanium) that was completely incompatible with anything made before and required large scale rewrites of the OS and software designs. Luckily, voices of reason (AMD / Linus) were there to counter with "just extend the address space and register count a little while keeping backwards compatibility" and amd64 (x86-64) was therefore able to smack the industry with reason. We didn't have to simultaneously support two completely different architectures at the same time. With Intel's new arch being completely optional, separate, and not compatible with existing software, there was very little incentive to switch. Meanwhile, the backwards compatibility of amd64 allowed software developers to instead upgrade at their own pace (or even not all) and an OS today can run a mix of 32 and 64 bit software and just work.

      IPv6 is the Itanium of networking. IPv4's biggest flaw was the lack of addresses. Like a 32 bit CPU only being able to handle 4 gigs of memory, IPv4 can only handle 4 billion or so addresses. Shamefully, nobody said "just extend the address space a little and keep backwards compatibility" in the network industry. If they would have, we would have switched long long ago and would have saved us so much time and heartache.

      • (Score: 2) by pe1rxq on Monday July 08, @10:25PM (1 child)

        by pe1rxq (844) on Monday July 08, @10:25PM (#1363496) Homepage

        'Just extend the address space a little and keep backwards compatibility' is nonsense. What do you even mean with 'backwards compatible' if legacy stacks can't reach half the world? Where would you hide the extra bits in a IPv4 header in a compatible way? And all this just to keep the format field at '4' instead of increasing it a bit like it was always intended to be used?
        Its not that nobody said this. Statements like this come up pretty much in every IPv6 discussion. And they are still just as silly as 20 years ago.

        • (Score: 2) by hendrikboom on Thursday July 25, @08:02PM

          by hendrikboom (1125) on Thursday July 25, @08:02PM (#1365678) Homepage Journal

          When IPv6 was introduced, there wasn't even a chosen standard embedding of IPv4 addresses into the IPv6 address space. No implicit interoperability was possible in either direction, not even with (incomplete) protocol conversion.
          That issue has since been resolved.

  • (Score: 4, Insightful) by darkfeline on Saturday July 06, @10:30PM

    by darkfeline (1030) on Saturday July 06, @10:30PM (#1363342) Homepage

    IPv6 is like Kubernetes or Bazel. It's great that it exists, but most people don't need it, and if they don't have it already, the migration costs likely aren't worth it. For the people who do need it, the migration costs are worth it.

    --
    Join the SDF Public Access UNIX System today!
(1)