Re: Guest Column: Kentik's Doug Madory, Last Call for Upcoming ISOC Course + More
Ryan Hamel wrote: > For you to say, "my privacy has been sold", is simply not true. I agree with you somewhat about tracking links. They only spy on a person when that person tries to follow them. I do find it much less useful to read mailing lists that include references to external resources that I decline to access, because I don't want to follow bugged links. But the "web bugs" that I mentioned as a second default-on Mailchimp tracking technology ARE specifically designed to be triggered any time a recipient reads a message in an HTML-based web browser. Back when postal mail was the default, senders had no idea whether the recipient opened, read, or forwarded a letter, versus tossing it into the fireplace as kindling. Society carried forward that expectation when postal mail was gradually replaced by electronic mail. Ordinary email senders don't know if you have read their message (unless they get social clues from your subsequent actions, just as with paper mail). Tracking was never part of the Internet email protocols; it was glued-on by abusing HTML email features and making unique URLs sent to each recipient, whose corresponding web server logs when they are accessed. These email tracking technologies deliberately violate the social expectation that reading a letter is a private act. They produce detailed records of the private, in-home or at-work activities of every recipient. They do all this covertly; you will not find a MailChimp mailing list message plainly telling you, "If you want to safeguard your privacy as an email reader, do not open these messages, because we have filled them with spyware." That would produce too many unsubscribes and too much outrage. Instead, a recipient has to be technically sophisticated to even notice that it's happening. (Many bulk email senders also don't know that their emails have spyware quietly inserted into them as they are distributed. I have engaged on this topic with many nonprofit CEOs and marketing executives, who really had no idea.) Those detailed email-reading and link-clicking records are not just accessible to the sender. There's an agency problem. They are kept and stored and sold by the intermediary (MailChimp), both individually and in bulk. They are accessible to any government that wants to ask, without a warrant, without probable cause, in bulk or individually, since they are "third-party" records about you, like your banking records or license-plate-reader records. They are accessible to private investigators via data brokers. They are accessible to any business that offers a sufficiently attractive deal to MailChimp -- places like Google or Facebook who make billions of dollars a year from tracking people to manipulate them with advertising. And wouldn't you like to know just which emails your competitors' engineers and executives are reading, and when, and where, and how many times, and whether they forwarded the messages? (I've often wanted the Google Detective Agency, that I could merely pay to tell me what my wife or my competitor or that rude guy who insulted me is searching for on Google, what web pages they are looking at, what emails they are reading or sending, and exactly where they are navigating in their car or on their bike or on transit. Google has all this information; why won't they sell it to me? They definitely sell it to the government, so why not to me? It's amazing to me that people treat Google like Santa Claus giving them free gifts, when it's really like an NSA.gov that is unencumbered by laws or oversight. MailChimp isn't as bad as Google. Its scope is smaller, but its defaults are deliberately bad, and it's created quite a honeypot of trillions of records about billions of people. The point is that besides being a gross violation of the personal privacy of the home and office, this data also has real commercial value. I suggest that as a technically aware organization, NANOG.org should not be creating detailed spy dossiers on its members who read emails, and then letting its subcontractor MailChimp sell or trade that info out into the world. John Gilmore
Re: Guest Column: Kentik's Doug Madory, Last Call for Upcoming ISOC Course + More
It is totally possible to turn off the spyware in MailChimp. You just need to buy an actual commercial account rather than using their "free" service. To save $13 or $20 per month, you are instead selling the privacy of every recipient of your emails. See: https://mailchimp.com/help/enable-and-view-click-tracking/ "Check the Track clicks box to enable click tracking, or uncheck the box to disable click tracking. ... Mailchimp will continue to redirect URLs for users with free account plans to protect against malicious links. ... When a paid user turns off click tracking, Mailchimp will continue to redirect their URLs until certain account activity thresholds are met." Don't forget to turn off the spyware 1x1 pixel "web bugs" that MailChimp inserts by default, too: https://mailchimp.com/help/about-open-tracking/ John
Re: NTP Sync Issue Across Tata (Europe)
Forrest Christian (List Account) wrote: > > > At some point, using publicly available NTP sources is redundant > > > unless one wants to mitigate away the risks behind failure of the GPS > > > system itself. On Fri, Aug 11, 2023, 3:33 AM Masataka Ohta wrote: > > Your assumption that public NTP servers were not GPS-derived NTP > > servers is just wrong. Subsequent conversation has shown that you are both right here. Yes, many public NTP servers ARE using GPS-derived time. Yes, some public NTP servers ARE NOT using GPS-derived time. Up to this point, popular public NTP pools have not made these distinctions readily configurable, though. For sites that need to run even during a war, or a similar situation that is likely to disrupt or distort GPS, they might like to have access to NTP servers that are completely independent from GPS. At one point I proposed that some big NTP server pools be segregated by names, to distinguish between GPS-derived time and national-standard derived time. For example, two domain names could be e.g.: fromnist.pool.tick.tock fromgps.pool.tick.tock If you wanted particular redundancy, say because you have a local GPS clock and you want a non-GPS check on that, you'd use fromnist.pool.tick.tock (or fromnict.pool.tick.tock for the Japanese timebase, etc). (If you were agnostic about where your times comes from, you would just use a generic domain name like vendorname.pool.tick.tock.) An automated tool could periodically trace that the stratum 0 source currently being used by each node in each of the pools is actually the same as advertised in its domain name. Alerting any difference to the relevant system administrators would allow those clocks to continue running with a backup timebase, while making it more likely that some human would work to restore their access to the correct stratum 0 source. So far this is just an idea. John PS: When we say "GPS", do we really mean any GNSS (global navigation satellite system)? There are now four such systems that have global coverage, plus regionals. While they attempt to coordinate their time-bases and reference-frames, they are using different hardware and systems, and are under different administration, so there are some differences in the clock values returned by each GNSS. These differences and discontinuties have ranged up to 100ns in normal operation, and higher amounts in the past. See: Nicolini, Luca; Caporali, Alessandro (9 January 2018). "Investigation on Reference Frames and Time Systems in Multi-GNSS". Remote Sensing. 10 (2): 80. doi:10.3390/rs10010080. https://www.mdpi.com/2072-4292/10/1/80
Re: NTP Sync Issue Across Tata (Europe)
> I was also speaking specifically about installing GPS antennas in > viable places, not using a facility-provided GPS or NTP service. Am I confused? Getting the time over a multi-gigabit Internet from a national time standard agency such as NIST (or your local country's equivalent) should produce far better accuracy and stability than relying on locally received GPS signals. GPS uses very weak radio signals which are regularly spoofed by all sorts of bad actors: https://www.gps.gov/spectrum/jamming/ for all sorts of reasons (like misleading drone navigation): https://en.wikipedia.org/wiki/Iran%E2%80%93U.S._RQ-170_incident Depending on satnav systems creates a large single point of failure for worldwide civilian infrastructure. Jamming GPS with subtly fake time data near big data centers seems like an easy move that would cause all sorts of distributed algorithms to start failing in unusual ways. And in a more serious wartime attack, many or most GPS satellites themselves would be destroyed or disabled. Yet digital radio modulations like FT8 or DMR rely on tight time synchronization among different transmitters. So do many modern cellphone modulations -- not to mention distributed database sync algorithms. Depending on any of these for emergency communications when their time comes from GPS, is a recipe for having no communications during wars or cyber-wars in which GPS satellites are attacked or jammed. See a longer explanation here: https://www.ardc.net/apply/grants/2020-grants/grant-ntpsec/ I suspect that even today, if you rely on civilian GPS time near the US White House, Pentagon, or other military targets like bases, you will discover "anomalies" in the local radio GPS data, compared to what you get from an authenticated time standard over NTP. How reliable is civilian GPS time in Ukraine these days? John
Re: Alternative Re: ipv4/25s and above Re: 202211201009.AYC
John Curran wrote: >> https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-240/ > > ... this Internet draft ... can't be safely deployed in the actual > real-world Internet The draft *has been* safely deployed in the actual real-world Internet. It is in all Linux nodes since 2008, and all recent Cisco routers. We know that this step is safe, because it was already done a decade ago, and the Internet did not descend into flames (except on mailing lists ;-). The draft is just trying to help the paperwork catch up with the implementations. So it seems that John Curran was criticizing a strawman rather than the draft above (which I co-authored). Perhaps the confusion is that John thought that the draft would actually allocate 240/4 addresses to end-users for end-user purposes. Doing that would be unsafe in today's big Internet (though major squatters are already using these addresses in private clouds, as RIPE and Google have documented). Allocating these addresses would be far more controversial than just updating the code in IPv4 implementations. Therefore, the draft doesn't do that. Instead it would just clear out the cobwebs in implementations, which would some years later, after more work, allow a responsible party to make such allocations. John's suggestion that "it's unsafe to do this safe change, because we don't yet know *when* some later change will be safe" is simply incorrect. Perhaps what the draft failed to explain is that "Merely implementing unicast treatment of 240/4 addresses in routers and operating systems, as this draft proposes, does not cause any interoperability issues. Hundreds of millions of IPv4 nodes currently contain this unicast treatment, and all are interoperating successfully with each other and with non-updated nodes." We'll add something like that to the next version. John Gilmore IPv4 Unicast Extensions Project
Re: any dangers of filtering every /24 on full internet table to preserve FIB space ?
Randy Bush wrote: > it is a tragedy that cidr and an open market has helped us more than > ipv6 has. True. Maybe cidr and an open market for ipv6 addresses would reduce the tragedy? John
Re: [External] Normal ARIN registration service fees for LRSA entrants after 31 Dec 2023
John Curran wrote: > [challenges by legacy registrants] has been before judges and resolved > numerous times. > > We’ve actually had the matter before many judges, and have never been > ordered to do anything other than operate the registry per the number > resource policy as developed by this community – this has been the > consistent outcome throughout both civil and bankruptcy > proceedings. Is there a public archive of these court proceedings? Or even a list of which cases have involved ARIN (or another RIR)? What can the community learn from what help resource holders have asked courts for, and what help they eventually got? John PS: Re another RIR: There's a short list of some of the ~50 lawsuits against AFRINIC in its Wikipedia page: https://en.wikipedia.org/wiki/AFRINIC#Controversies_&_Scandals These are mostly to do with corruption, theft, and harassment. But an important subtheme includes what power AFRINIC has to seize IP addresses that were legitimately allocated to recipients. In the "Cloud Innovation" case, CI got addresses under standard policy, but years later, as I recall, AFRINIC tried to retroactively impose a new "no renting vm's" policy and a new requirement that all the addresses be hosted in Africa, even by global customers. After AFRINIC threatened to immediately revoke CI's membership and take back the addresses over this, CI sued AFRINIC to keep the status quo, keeping their business alive, while the courts sort out whether AFRINIC has the power to do so. Since then, it's mostly been procedural scuffling and some bad faith negotiations. If neither party goes bankrupt nor settles, it's possible that the courts of Mauritius will answer the question about whether their RIR has the power to impose new policies and then reclaim allocated addresses for violating them.
Re: [External] Normal ARIN registration service fees for LRSA entrants after 31 Dec 2023 (was: Fwd: [arin-announce] Availability of the Legacy Fee Cap for New LRSA Entrants Ending as of 31 December 20
John Curran wrote: > ... the long-term direction is to provide the same services to all > customers under the same agreement and fees – anything else wouldn’t > be equitable. There are many "anything else"s that would indeed be equitable. It is equitable for businesses to sell yesterday's bread at a lower price than today's bread. Or to rent unused hotel rooms to late-night transients for lower prices than those charged to people who want pre-booked certainty about their overnight shelter. ARIN could equitably charge different prices to people in different situations; it already does. And ARIN could equitably offer services to non-members, by charging them transaction fees for services rendered, rather than trying to force them into a disadvantageous long term contract. Please don't confuse "seeking equity" with "forcing everyone into the same procrustean bed". As a simple example, ARIN's contract need not require its customers to give up their resources when ceasing to pay ARIN for services. (There's an existence proof: RIPE's doesn't.) Such a contract would likely result in more-equitable sharing of costs, since it would encourage legacy holders to pay ARIN (and legacy holders are still more than a quarter of the total IP addresses, possibly much more). The fact that ARIN hasn't made this happen says nothing about equity; it's about something else. This whole tussle is about power. ARIN wants the power to take away legacy resources, while their current owners don't want that to happen. ARIN wants to be the puppeteer who pulls all the strings for the North American Internet. It pursues this desire by stealth and misdirection (e.g. "We strongly encourage all legacy resource holders who have not yet signed an LRSA to cover their legacy resources to consider doing so before 31 December 2023 in order to secure the most favorable fees for their ARIN Services...") ARIN is also trying to encourage ISPs to demand RPKI before providing transit to IP address holders, which would turn its optional RPKI service (that it has tied by contract into ARIN gaining control over legacy resources) into an effectively mandatory RPKI service. ARIN hides its power grab behind "our policies are set by our community" and "our board is elected by our community" misdirections. Its voting community consists almost entirely of those who aren't legacy holders (by definition: if you accept their contract, your legacy resource ownership goes away; if you don't, you can't vote). That community would love to confiscate some "underused" legacy IP addresses to be handed out for free to their own "waiting list". So this is equivalent to putting foxes in charge of policy for a henhouse. Now that markets exist for IP addresses, all that IP addresses need is a deed-registry to discourage fraud, like a county real-estate registrar's office. IP addresses no longer need a bureacracy for socialistic determinations about which supplicants "deserve" addresses. Addresses now have prices, and if you want some, you buy them. Deed registries get to charge fees for transactions, but they don't get to take away your property, nor tell you that you can't buy any more property because they disapprove of how you managed your previous properties. Actual ownership of real estate is defined by contracts and courts, not by the registry, which is just a set of pointers to help people figure out the history and current status of each parcel. The registry is important, but it's not definitive. Deed-registry is apparently not a model that ARIN wants to be operating in. They initially tried to refuse to record purchases of address blocks, because it violated their model of "if you don't use your IP addresses, you must give them back to us and receive no money for them". They saw their job as being the power broker who hands out free favors. But when their supply of free IP addresses dried up, they had no remaining function other than to record ownership (be a deed registry), and to run an occasional conference. It dawned on them that if they refused to record these transactions, they would not even be a reliable deed-registry; they would have entirely outlived their usefulness. So they reluctantly agreed to do that job, but their policies are still left over from their power-broker past. They'd love to go back to it, if only they could figure out how. IPv6? Sure! RPKI maybe? Worth a try! ARIN prefers to be a power broker rather than a scribe. Who can blame them for that? But don't mistake their strategy for stewardship. "Doing what the community wants" or "seeking the equitable thing" quacks like stewardship, so of course they brand themselves that way. But in my opinion their power-seeking is self-serving, not community-serving. John Gilmore
Re: Normal ARIN registration service fees for LRSA entrants after 31 Dec 2023 (was: Fwd: [arin-announce] Availability of the Legacy Fee Cap for New LRSA Entrants Ending as of 31 December 2023)
John Curran wrote: > > We strongly encourage all legacy resource holders who have not yet > > signed an LRSA to cover their legacy resources to Randy Bush wrote: > consult a competent lawyer before signing an LRSA Amen to that. ARIN's stance on legacy resources has traditionally been that ARIN would prefer to charge you annually for them, and then "recover" them (take them away from you) if you ever stop paying, or if they ever decide that you are not using them wisely. If you once agree to an ARIN contract, your resources lose their "legacy" status and you become just another sharecropper subject to ARIN's future benevolence or lack thereof. The change recently announced by John Curran will make the situation very slightly worse, by making ARIN's annual fees for legacy resources changeable at their option, instead of being capped by contract. ARIN management could have changed their offer to be better, if they wanted to attract legacy users, but they made an explicit choice to do the opposite. By contrast, RIPE has developed a much more welcoming stance on legacy resources, including: * retaining the legacy status of resources after a transfer or sale * allowing resources to be registered without paying annual fees to RIPE (merely paying a one-time transaction fee), so that later non-payment of annual fees can't be used as an excuse to steal the resources. * agreeing that RIPE members will keep all their legacy resources even if they later cease to be RIPE members You are within the RIPE service area if your network touches Europe, northern Asia, or Greenland. This can be as simple as having a rented or donated server located in Europe, or as complicated as running a worldwide service provider. If you have a presence there, you can transfer your worldwide resources out from under ARIN policies and put them under RIPE's jurisdiction instead. Moving to RIPE is not an unalloyed good; Europeans invented bureaucracy, and RIPE pursues it with vigor. And getting the above treatment may require firmly asserting to RIPE that you want it, rather than accepting the defaults. But their motives are more benevolent than ARIN's toward legacy resource holders; RIPE honestly seems to want to gather in legacy resource holders, either as RIPE members or not, without reducing any of the holders' rights or abilities. I commend them for that. Other RIRs may have other good or bad policies about legacy resource holders. As Randy proposed, consult a lawyer competent in legacy domain registration issues before making any changes. John
Re: Serious Juniper Hardware EoL Announcements
Matthew Petach wrote: > https://cacm.acm.org/news/257742-german-factory-fire-could-worsen-global-chip-shortage/fulltext > > That was the *sole* supplier of extreme ultraviolet lithography machines > for every major chip manufacturer on the planet. > > Chip shortages will only get worse for the next several years. The light > at the end of the tunnel is unfortunately *not* coming from an ultraviolet > lithography machine. :( It's quite trendy (but inaccurate) to declare that everything sucks, human life on the planet is ending, etc. Matthew's last paragraph seems to be one of those unduly dire conclusions, based on subsequent news after January. See: https://www.asml.com/en/news/press-releases/2022/update-fire-incident-at-asml-berlin https://www.cnbc.com/2022/01/19/asml-profit-beats-despite-berlin-fire-sees-20percent-sales-growth-in-2022-.html Those with a detailed interest in the topic can speak directly with Monique Mols, head of media relations at ASML.com, at +31 652 844 418, or Ryan Young, US media relations manager, +1 480 205 8659. Ryan confirmed to me today that the latest news is in the above links: there is expected to be no impact from the fire on ASML's extreme UV delivery schedule. He says they will provide a further update at their large annual meeting in about a month. John
Re: Serious Juniper Hardware EoL Announcements
Dave Taht wrote: > > Then it was "what can we do with what we can afford" now it's more > > like "What can we do with what we have (or can actually get)"? > > Like, working on better software... Like, deploying the other 300 million IPv4 addresses that are currently lying around unused. They remain formally unused due to three interlocking supply chain problems: at IETF, ICANN, and vendors. IETF's is caused by a "we must force everyone to abandon trailing edge technology" attitude. ICANN's is because nobody is sure how to allocate ~$15B worth of end-user value into a calcified IP address market dominated by government-created regional monopolies doing allocation by fiat. Vendors have leapfrogged the IETF and ICANN processes, and most have deployed the key one-line software patches needed to fully enable these addresses in OS's and routers. Microsoft is the only major vendor seemingly committed to never doing so. Our project continues to track progress in this area, and test and document compatability. John IPv4 Unicast Extensions Project
Re: 2749 routes AT RISK - Re: TIMELY/IMPORTANT - Approximately 40 hours until potentially significant routing changes (re: Retirement of ARIN Non-Authenticated IRR scheduled for 4 April 2022)
Job Snijders via NANOG wrote: > our community also has to be cognizant about there being parts of the > Internet which are not squatting on anyone's numbers *and* also are > not contracted to a specific RIR. Let's not undermine one of the few remaining widely distributed (with no center) technical achievements behind the Internet -- the decentralized routing system. I'm on the board of a large legacy allocation that is deliberately NOT an ARIN (or other RIR) member. And I have a small address block of my own, ditto. ARIN doesn't provide authenticated RPKI entries for just anybody. You have to pay them for that service. And in order to pay them, you have to sign their contract. And if you sign that contract, ARIN can take away your legacy allocation -- anytime they decide it would be in their best interest. Whereas, if you don't sign, the courts have held that you have a *property right* in your IP addresses and they *belong* to you. As a result, most legacy address holders (a large fraction of the Internet addresses) have declined to sign such contracts, pay such bills, and thus can't be in the ARIN authenticated routing registry. For years, ARIN has been deliberately limiting access to the RPKI registry as a lever to force people to sign one-sided contracts beneficial to ARIN. (They do the same lever thing when you sell an address block -- at ARIN, it loses its legacy status, requiring the recipient to pay annual rent to ARIN, and risk losing their block if political winds shift.) The pro-RPKI faction also seems to have completely ignored what I consider a major concern among anti-RPKI folks. The distributed Internet routing system is resilient to centralized failures, and should remain so. Inserting five points of failure (signatures of RIRs) would undermine that resilience. Also, centralizing control over route acceptance can be used for censorship. If the RIRs succeed in convincing "enough of the net" to reject any route that doesn't come with an RIR signature, then any government with jurisdiction over those RIRs can force them to not sign routes for sites that are politically incorrect. How convenient -- for authoritarians. You can have all the IP addresses you want, you just can't get 90% of the ISPs in the world to route packets to them. There is no shortage of Horsemen of the Infopocalypse (child porn, terrorism, sex slavery, Covid misinformation, manipulative propaganda, war news, copyright violations, etc, etc, etc) that Absolutely Need To Be Stamped Out Today whenever politicians decide that Something Must Be Done. As an example, we have regularly seen courts force centralized domain registrars to reject perfectly good applicants for just such reasons (e.g. SciHub). The distributed Internet has "routed around" their ability to censor such information via the routing table. ISPs should not hand governments a tool that they have abused so many times in the past. John
Re: Let's Focus on Moving Forward Re: V6 still not supported
Tom Beecher wrote: > I'd be curious to see the data you guys have collected on what it has been > confirmed to work on if that's available somewhere. The Implementation Status of unicast 240/4 is in the Appendix of our draft: https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-240/ Enjoy -- it's a long list. In addition to what's in the published draft, some research that I did last night also determined that 240/4 as unicast was released in Fedora 9 in May 2008, and in Ubuntu 8.10 in October 2008. John
Re: Let's Focus on Moving Forward Re: V6 still not supported
Christopher Morrow wrote: > I think the advice in the draft, and on the quoted page of Google cloud > docs is that you can use whatever address space you want for your voc > network. I think it also says that choosing poorly could make portions if > the internet unreachable. > > I don't see that the docs specify usage of 240/4 though. Thank you for catching this! Draft-schoen-intarea-unicast-240 links its reference [VPC] to: https://cloud.google.com/vpc/docs/vpc#valid-ranges As late as March 18, 2022, that page included a table of "Valid ranges" that include "240.0.0.0/4", which you can see in the Internet Archive's "Wayback Machine" copy of the page: http://web.archive.org/web/20220318080636/https://cloud.google.com/vpc/docs/vpc A subnet's primary and secondary IP address ranges are regional internal IP addresses. The following table describes valid ranges. ... 240.0.0.0/4 Reserved for future use (Class E) as noted in RFC 5735 and RFC 1112. Some operating systems do not support the use of this range, so verify that your OS supports it before creating subnets that use this range. However, as of March 20, Google moved that table into a subsidiary page, the "Subnets overview", which is now linked from the original page: http://web.archive.org/web/20220320031522/https://cloud.google.com/vpc/docs/vpc http://web.archive.org/web/20220328102630/https://cloud.google.com/vpc/docs/subnets The same information about 240/4 is there. Thanks again for the bug report! We'll update the URL in the next version of the draft. John
Re: Let's Focus on Moving Forward Re: V6 still not supported
Tom Beecher wrote: > > */writing/* and */deploying/* the code that will allow the use of 240/4 the > > way you expect > > While Mr. Chen may have considered that, he has repeatedly hand waved that > it's 'not that big a deal.', so I don't think he adequately grasps the > scale of that challenge. >From multiple years of patching and testing, the IPv4 Unicast Extensions Project knows that 240/4 ALREADY WORKS in a large fraction of the Internet. Including all the Linux servers and desktops, all the Android phones and tablets, all the MacOS machines, all the iOS phones, many of the home wifi gateways. All the Ethernet switches. And some less popular stuff like routers from Cisco, Juniper, and OpenWRT. Most of these started working A DECADE AGO. If others grasp the scale of the challenge better than we do, I'm happy to learn from them. A traceroute from my machine to 240.1.2.3 goes through six routers at my ISP before stopping (probably at the first default-route-free router). Today Google is documenting to its cloud customers that they should use 240/4 for internal networks. (Read draft-schoen-intarea-unicast-240 for the citation.) We have received inquiries from two other huge Internet companies, which are investigating or already using 240/4 as private IPv4 address space. In short, we are actually making it work, and writing a spec for what already works. Our detractors are arguing: not that it doesn't work, but that we should instead seek to accomplish somebody else's goals. John PS: Mr. Abraham Chen's effort is not related to ours. Our drafts are agnostic about what 240/4 should be used for after we enable it as ordinary unicast. His EzIP overlay network effort is one that I don't fully understand. What I do understand is that since his effort uses 240/4 addresses as the outer addresses in IPv4 packets, it couldn't work without reaching our goal first: allowing any site on the Internet to send unicast packets to or from 240.0.0.1 and having them arrive.
Re: V6 still not supported
Pascal Thubert \(pthubert\) via NANOG wrote: > I'm personally fond of the IP-in-IP variation that filed in 20+ years > ago as US patent 7,356,031. No wonder -- you are listed as the co-inventor! Just the fact that it is patented (and the patent is still unexpired) would make it a disfavored candidate for an Internet transition technology. It was not nice of y'all to try to get a monopoly over nesting headers for making an overlay network that tunnels things to distant routers. You have to certify that your work is original in order to even apply for a patent. So, nobody had ever thought of that before y'all did? Really? John
Re: BOOTP & ARP history
Michael Thomas wrote: > There were tons of things that were slapped onto IP that were basically > experimental like ARP and bootp. CIDR didn't even exist back then. Speaking as one of the co-designers of BOOTP (RFC 951): yes, it was experimental. So why was it "slapped onto" IP? Well, in those days the IP protocol itself (now known as IPv4) was an experiment. The previous method of configuring each computer that was plugged into an IP network, was to have a human-to-human conversation with your network administrator, who would manually write down your new IP address on the local network, and also tell you the IP address of a local gateway. You would type this into a file in /etc and save it on your new computer's local hard drive, and your experimental network would all start working. Then it became 1985, disk drives were small and expensive, flash memory nonexistent. As a cost-reduction measure, Sun built some of the first end-user computers that could operate without any nonvolatile local storage: diskless workstations. They needed a way to boot without having a system administrator (or end user) type in the machine's IP address and the address of their gateway every time it booted. We had a network, why didn't we communicate this over the network? Sun designed and implemented a way of doing this, called RARP (RFC 903). This did not use IP packets; it required accessing the local Ethernet using packets with a unique Ethertype. And it only worked on Ethernet, not on any other possible LAN technology. I was the maintainer of the bootstrap ROMs in Sun workstations. Bill Croft and I thought we could do better. We put our heads together and said, "why can't we use IP broadcast packets for this?" (IP support for broadcast packets was also an experiment at around that time. IP multicast was barely a cloud on the horizon.) You could even write a portable UNIX program to be the BOOTP server or client (unlike for RARP). We got the BOOTP protocol working in prototype. Bill understood the RFC submission process, and we got it published as an experimental RFC. Sun didn't care. Their products shipped using RARP. Bill and I had lots of other things to do, so we ignored BOOTP for years. But Sun had a run of good luck at supporting and creating industry standards like TCP/IP and NFS. 3000 miles away, some competitors hated Sun's success, so they decided to standardize their products on "anything that doesn't work like Sun's". This was the UNIX wars, which distracted all the UNIX vendors. They should've been watching Microsoft, which proceeded to steamroller the computing market and make almost all of them irrelevant. I think it might have been DEC that first adopted BOOTP for actually bootstrapping their products, and others followed suit. I don't actually know why they picked it -- especially because I was co-author and I worked at Sun. But they started using it, and it caught on among that consortium of non-Sun UNIX vendors. And eventually, Mitch Bradley, a hardware designer at Sun with a penchant for FORTH programming, built the boot code for the first SPARC machines, and implemented BOOTP in it, so even Sun started using it. By 1993, others had built DHCP (RFC 1541) on top of BOOTP, and that went onto the standards track at IETF. Everybody who liked the old way was free to continue manually typing human-assigned IP addresses into every new computer on their network and keeping them in local nonvolatile storage. Plugging into an Ethernet used to require inserting an electric drill into a fat yellow coaxial cable in your ceiling (see https://en.wikipedia.org/wiki/10BASE5), and then cleaning out the copper detritis that would short the cable and bring down the whole network, and attaching in a separate transceiver box with a "sting tap" that would touch the center conductor of the coax, then running a separate transceiver cable to your computer. The manual IP address configuration was only a small portion of the pain. But 3Com migrated Ethernet to twist-on BNC connectors (https://en.wikipedia.org/wiki/10BASE2), and users started migrating toward products that worked without bringing in a technician. BOOTP and DHCP offered a "plug and play" experience that those users craved. So what started as experiments eventually became expected parts of the IP standards. ARP was "slapped on" in 1982, long before RARP or BOOTP. The original IP specs required that the LAN address must fit into the low order bits of your IP netblock. This wasn't well thought through, but IP was an experiment and there were very few other experiments for its designers to learn from. It worked ok when ARPANET was your LAN (see the original use of 10/8), and when everybody else had Class A addresses, like the packet radio network or 3-megabit Experimental Ethernet users. But it didn't scale up, and it didn't work at all for 10-megabit industry standard Ethernet, with 48-bit addresses much longer than IP
Re: V6 still not supported
> > Let me say that again. Among all the reasons why IPv6 didn't take > > over the world, NONE of them is "because we spent all our time > > improving IPv4 standards instead". > > I'll somewhat call bullshit on this conclusion from the data > available. True, none of the reasons directly claim "IPv6 isn't good > enough because we did X for v4 instead", yet all of them in some way > refer back to "insufficient resources to make this the top priority." > which means that any resources being dedicated to improving (or more > accurately further band-aiding) IPv4 are effectively being taken away > from solving the problems that exist with IPv6 pretty much by > definition. Hi, Owen. Your reasoning proves too much. You propose that every minute of every day that every human isn't actively working at top priority to make IPv6 the default protocol on the Internet, are misguided efforts. "Pretty much by definition." "Any resources being dedicated to" eating, sleeping, going to the bathroom, listening to music, painting a canvas, repairing cars, steering ships, growing food, running railroads, going to school, going to work, riding bicycles, ending homelessness, stopping wars, reforming drug laws, band-aiding IPv4, reducing corruption in government, posting to mailing lists (as you pointed out -- by posting a message to a mailing list!), hopping, skipping, and jumping, "are effectively being taken away from solving the problems that exist with IPv6." Given the billions of people who eat and sleep for HOURS every day, I think I am doing pretty well by just coordinating three people part-time trying to improve IPv4 a little bit. The eaters' and sleepers' level of non-IPv6 effort is billions of times stronger than my level of non-IPv6 effort. Can you forgive me? John
Re: V6 still not supported
It is great to see NANOG members describing some of the real barriers to widespread IPv6 deployment. Buggy implementations, lack of consumer demand, too many other things to do (like rapidly deploying fiber to customers before they switch to a competitor), lack of IPv6 expertise at ISPs, lack of ISP demand driving lack of supplier support, and doubled testing and qualification workload. As Tim Howe wrote: >... I do not really blame those who don't, because in order > to get where we are I had to make it my personal mission in life to get > to a passive FTTP configuration that would work with functional parity > between v4 and v6... > For over a year I had to test gear, which requires a lot of > time and effort and study and support and managerial latitude. I had > to isolate bugs and spend the time reporting them, which often means > making a pain in the butt out of yourself and championing the issue > with the vendor (sometimes it means committing to buying things). I > had to INSIST on support from vendors and refuse to buy things that > didn't work. I had to buy new gear I would not have otherwise needed. > I also had to "fire" a couple of vendors and purge them from my > network; I even sent back an entire shipment of gear to a vendor due to > broken promises. > Basically I had to be extremely unreasonable. My position is > unique in that I was able to do these things and get away with it. I > can't blame anyone for not going down that road. What struck me is how NONE of those challenges in doing IPv6 deployment in the field had anything to do with fending off attempts to make IPv4 better. Let me say that again. Among all the reasons why IPv6 didn't take over the world, NONE of them is "because we spent all our time improving IPv4 standards instead". John Gilmore
Re: CC: s to Non List Members (was Re: 202203080924.AYC Re: 202203071610.AYC Re: Making Use of 240/4 NetBlock)
John Levine wrote: > FWIW, I also don't think that repurposing 240/4 is a good idea. To be > useful it would require that every host on the Internet update its > network stack, which would take on the order of a decade... Those network stacks were updated for 240/4 in 2008-2009 -- a decade ago. See the Implementation Status section of our draft: https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-240/ Major networks are already squatting on the space internally, because they tried it and it works. We have running code. The future is now. We are ready to update the standards. The only major OS that doesn't support 240/4 is Microsoft Windows -- and it comes with regular online updates. So if IETF made the decision to make it unicast space, most MS OS users could be updated within less than a year. > It's basically > the same amount of work as getting everything to work on IPv6. If that was true, we'd be living in the IPv6 heaven now. It doesn't take any OS upgrades for "getting everything to work on IPv6". All the OS's and routers have supported IPv6 for more than a decade. Whatever the IPv6 transition might require, it isn't comparable to the small effort needed to upgrade a few laggard OS's to support 240/4 and to do some de-bogonization in the global Internet, akin to what CloudFlare did for 1.1.1.1. John
Re: VPN recommendations?
Mike Lyon wrote: > How about running ZeroTier on those Linux boxes and call it a day? > https://www.zerotier.com/ ZeroTier is not a free-as-in-freedom project. Running it in Linux boxes or network appliances to provide a VPN to paying customers may be prohibited (at least for some customers, and before 2025) by its convoluted license: https://github.com/zerotier/ZeroTierOne/blob/master/LICENSE.txt I recommend using something that doesn't have litigious companies nitpicking about what you can and can't use it for. John Gilmore
Re: questions about ARIN ipv6 allocation
Owen DeLong via NANOG wrote: > The double billing (had it been present at the time) would have prevented me > from signing the LRSA for my IPv4 resources. Owen, the root of your problem is that you signed an LRSA with ARIN, rather than keeping your legacy resources un-tainted by an ARIN contract that deliberately reduced your rights. When ARDC transferred 44.192/10 via ARIN, the recipient lost the legacy status of the address block. That was an ARIN requirement, which was OK with that particular recipient. However, ARIN is not your only option. It is possible to transfer legacy resources such as IPv4 address blocks from ARIN to RIPE, having them be recognized as legacy blocks under RIPE jurisdiction. You can do this without signing any long term contract with RIPE, if you like; or you can choose to become a long-term paying RIPE member, under their fee schedule. All you need is to have any Internet resources in Europe -- like a virtual machine in a data center there, or a DNS server. I'm sure of this because I have done it; see https://apps.db.ripe.net/db-web-ui/lookup?source=ripe=209.16.159.0%20-%20209.16.159.255=inetnum The short-term contract for the transfer honors and retains the legacy status of those resources: that you own them, not the ARIN fiction that an RIR now controls them and will steal them from you if you stop paying them annually. Randy Bush detailed a similar transfer process back in 2016: https://archive.psg.com/160524.ripe-transfer.pdf The process is more bureaucratic and cumbersome than you expect; Europeans named bureacracy in the 1800s, and RIPE has raised it to a painful art. But once it's done, you are out from under the ARIN anti-legacy mentality forever. John Gilmore PS: If you want RPKI, which I didn't, you can sign a RIPE long term contract, pay them annually, and (according to Randy) they will STILL honor your ownership of your resources, unlike ARIN.
Re: private 5G networks?
Michael Thomas wrote: > > What do you mean 3rd Tier? > General Authorized Access? Taken from some random site looking it up. https://en.wikipedia.org/wiki/Citizens_Broadband_Radio_Service it has 3 tiers: * Incumbent access, primarily government and military radars, plus some pre-existing band users. * 3550 to 3650 MHz in 10MHz chunks, allocated for priority users by census tracts for up to 3 years, with up to 7 Priority Access Licenses per tract. Competitive bidding for getting these licenses. * General Authorized Access users can use any of those chunks that aren't assigned for priority use, or that nobody currently is transmitting on, plus another 50 MHz at 3650-3700 in free-for-all mode unless there are incumbents. A local Spectrum Access System (SAS) would program the individual devices to stay within the restrictions specified by the FCC and any licenses issued to the operator, for a particular geography. John PS: The CBRS radio devices can't turn on their transmitter until they talk a detailed negotiation to their SAS, via HTTP over TLS 1.2 over IPv4. IPv6 support is optional. None of this negotiation appears to happen over the radio, it's all apparently on Ethernet (or assumes some separate Internet provisioning not done in CBRS spectrum). And there's no discovery procedure, it's all done by manual configuration. See: https://winnf.memberclicks.net/assets/CBRS/WINNF-TS-0016.pdf
Re: Class E addresses? 240/4 history
Eliot Lear wrote: > In 2008, Vince Fuller, Dave Meyer, and I put together > draft-fuller-240space, and we presented it to the IETF. There were > definitely people who thought we should just try to get to v6, but > what really stopped us was a point that Dave Thaler made: unintended > impact on non-participating devices, and in particular CPE/consumer > firewall gear, and at the time there were serious concerns about some > endpoint systems as well. I was not in this part of IETF in those days, so I did not participate in those discussions. But I later read them on the archived mailing list, and reached out by email to Dave Thaler for more details about his concerns. He responded with the same general issues (and a request that we and everyone else spend more time on IPv6). I asked in a subsequent message for any details he has about such products that he thought would fail. He was unable or unwilling to point out even a single operating system, Internet node type, or firewall product that would fail unsafely if it saw packets from the 240/4 range. As documented in our Internet-Draft, all such products known to us either accept those packets as unicast traffic, or reject such packets and do not let them through. None crashes, reboots, fills logfiles with endless messages, falls on the floor, or otherwise fails. No known firewall is letting 240/4 packets through on the theory that it's perfectly safe because every end-system will discard them. As far as I can tell, what Eliot says really stopped this proposal in 2008 was Dave's hand-wave of *potential* concern, not an actual documented problem with the proposal. If anyone knows an *actual* documented problem with 240/4 packets, please tell us! (And as I pointed out subsequently to Dave, if any nodes currently in service would *actually* crash if they received a 240/4 packet, that's a critical denial of service issue. For reasons completely independent from our proposal, those machines should be rapidly identified and patched, rather than remaining vulnerable from 2008 thru 2021 and beyond. It would be trivial for an attacker to send such packets-of-death from any Linux, Solaris, Android, MacOS, or iOS machine that they've broken into on the local LAN. And even Windows machines may have ways to send raw Ethernet packets that could be crafted by an attacker to appear to be deadly IPv4 240/4 packets.) John
FreeBSD users of 127/8
J. Hellenthal wrote: > FreeBSD operators have been using this space for quite a long time for > many NAT'ing reasons including firewalls and other services behind > them for jail routing and such. > > https://dan.langille.org/2013/12/29/freebsd-jails-on-non-routable-ip-addresses/ > > That's just one example that I've seen repeated in multiple other > ways. One of which a jail operator with about 250 addresses out of > that range that enabled his jail routed services. Thank you for letting us know! We would be happy to improve the draft so that it has less impact on such pre-existing users. When we surveyed publicly visible applications based on Linux, we only found them configured to use the lowest /16. It's true that any system operator could configure their system in any part of 127/8, but we focused on the default configurations of popular software (such as systemd and Kubernetes). Do you know of any FreeBSD software that comes with a default configuration in 127/8 but not in 127/16? (It looks like the web page you referenced is about specific manual configuration, not about the default behavior of supplied software.) I do not know the details of FreeBSD jail configuration, nor the precise behavior of its loopback interface. From my limited understanding, it looks like the jail configured in the web page you referenced, with address 127.1.0.128/32 on lo1, would provide loopback service regardless of whether the default address on lo0 was 127.0.0.1/8 or 127.0.0.1/16. That's because lo1 is a separate interface from lo0, and the "lo" interfaces always loop back any packets sent through them, no matter what addresses are configured on them. (Indeed the example configures it with a 10.80.0.128 address as well, which would not normally be considered a loopback address.) So, if I am right, then even if our current Internet-Draft became a standard and FreeBSD was modified to implement it, the recommended commands would continue to work. The only impact would be that such a FreeBSD machine would be unable to reach a potential global Internet service hosted out on the Internet at address 127.1.0.128 (because a local interface has been configured at that address, shadowing the globally reachable address). I anticipate that no such global services would be created before 2026 at the very earliest (other than for reachability testing), and likely much later in the 2020's or early 2030's. If it turns out that FreeBSD usage of 127.1/16 is widespread, and the above analysis is incorrect or unacceptable to the FreeBSD community, we would be happy to modify the draft to retain default loopback behavior on 127.0.0.1/17 rather than 127.0.0.1/16. That would include both 127.0.x.y and 127.1.x.y as default loopback addresses. This would completely resolve the issue presented on the "FreeBSD jails on non-routable IP addresses" web page, while still recovering more than 16 million addresses for global use. The worst case might be if FreeBSD sysadmins have become accustomed to picking "random" addresses manually from all over the 127/8 space. If so, it is not unreasonable to expect that when manually configuring a node to use "non-routable" addresses, that in the passage of time, some of them might become routable in the future. When upgrading any machine to a new OS release, various small things typically need adjusting to fit into the revised OS. Renumbering the in-system use of up to a few hundred non-routable addresses like 127.44.22.66 into addresses like 127.0.22.66 (in a smaller non-routable range that still would still contain 65,000 or 130,000 addresses) might be one of those things that could be easily adjusted during such an upgrade. John
Re: Redeploying most of 127/8, 0/8, 240/4 and *.0 as unicast
David Conrad wrote: > Doesn't this presume the redeployed addresses would be allocated > via a market rather than via the RIRs? > > If so, who would receive the money? You ask great questions. The community can and should do the engineering to extend the IP implementations. If that doesn't happen, the rest of the issues are moot. There would be no addresses to allocate and no money to receive. It would take multiple years post-RFC and post-implementation before anyone could even contemplate doing allocation, of any sort. So while it's an entertaining sideline to think about later, at this point it is a distraction. John PS: It's conceivable that RIRs could allocate via a market. One has already done so (APNIC).
Re: Redeploying most of 127/8, 0/8, 240/4 and *.0 as unicast
Fred Baker wrote: > I tend to think that if we can somehow bless a prefix and make be > global unicast address space, it needs to become Global Unicast > Address Space. Yes, I agree. The intention is that with the passage of time, each prefix becomes more and more reachable, til it's as close to 100% as any other IP address. I was just suggesting a side point, that some kinds of IP users may be able to make earlier use of measurably less-reachable addresses. That could possibly enable them to get those addresses at lower prices (and with no guarantees), compared to people who want and expect 100% reachable IP addresses. Having users making actual early use of them would also encourage those users to actively work to improve their reachability, rather than passively waiting til "somebody else" improves their reachability. (Indeed, some adventurous early adopters might buy such addresses, actively improve their reachability, then 'flip' them for higher prices, as some people do with real-estate.) Just pointing out a side chance for a win-win situation. Most users would wait til surveys show high reachability. John
Re: Redploying most of 127/8 as unicast public
=?utf-8?B?TcOlbnM=?= Nilsson wrote: > The only viable future is to convert [to IPv6]. This is not > group-think, it is simple math. OK. And in the long run, we are all dead. That is not group-think, it is simple math. Yet that's not a good argument for deciding not to improve our lives today. Nor to fail to improve them for tomorrow, in case we live til then. John
Re: Redeploying most of 127/8, 0/8, 240/4 and *.0 as unicast
Nick Hilliard wrote: >> consider three hosts on a broadcast domain: A, B and >> C. A uses the lowest address, B accepts a lowest address, but C does >> not. Then A can talk to B, B can talk to C, but C cannot talk to A. >> This does not seem to be addressed in the draft. Section 3.4. Compatibility and Interoperability. Many deployed systems follow older Internet standards in not allowing the lowest address in a network to be assigned or used as a source or destination address. Assigning this address to a host may thus make it inaccessible by some devices on its local network segment. [there's more...] If you think that section needs improving, please send suggested text. We're happy to explain the implications better. Joe Maimon wrote: > its a local support issue only. That's also true. The only issues arise between your devices, on your LAN. Everybody else on the Internet is unaffected and can reach all your devices, including the lowest if your LAN uses it. Nothing forces you to use your lowest address, and we recommend that DHCP servers be configured by default to not to hand them out (no change from how they historically have been configured). We submitted a 6-line patch to the Busybox DHCP implementation in February to avoid hardcoded prevention of issuing a .0 or .255 address (which was wrong anyway, even without our proposal). The default in the config file continues to use a range excluding .0. The patch was merged upstream without trouble. See: https://github.com/schoen/unicast-extensions/blob/master/merged-patches/busybox/0001-Don-t-hardcode-refusing-to-issue-.0-or-.255-addresse.patch John
Re: Redeploying most of 127/8, 0/8, 240/4 and *.0 as unicast
Joe Maimon wrote: > And all thats needed to be done is to drop this ridiculous .0 for > broadcast compatibility from standards.why is this even controversial? Not to put words in his mouth, but that's how original BSD maintainer Mike Karels seemed to feel when we raised this issue for FreeBSD. He was like, what? We're wasting an address per subnet for 4.2BSD compatability? Since 1986? Let's fix that. John
Re: Redeploying most of 127/8, 0/8, 240/4 and *.0 as unicast
Randy Bush wrote: > as a measurement kinda person, i wonder if anyone has looked at how much > progress has been made on getting hard coded dependencies on D, E, 127, > ... out of the firmware in all networked devices. The drafts each have an Implementation Status section that describes what we know. The authors would be happy to receive updates for any of that information. 240/4 is widespread everywhere except in Microsoft products. It works so reliably that both Google and Amazon appear to be bootlegging it on their internal networks. Google even advises customers to use it: https://cloud.google.com/vpc/docs/vpc#valid-ranges 0/8, and the lowest address per subnet, have interoperable implementations that don't cause problems with legacy equipment, but they are not widely deployed. 0/8 has a few years in Linux & Android kernels and distros. Lowest address is in the most recent Linux and FreeBSD kernels, but not yet in any OS distros. In particular, OpenWRT doesn't implement it yet, which could easily be a fast path to a free extra IP address for anyone who has a compatible home router and a globally routed small network like a /29. We used RIPE Atlas to test the reachability of many addresses that end in .0 and .0.0, and they were reachable from all but one probe that we tried. Amazon offers https://ec2-reachability.amazonaws.com/ for this purpose; try it on your own network! Some embedded TCP stacks treat 127/8 as unicast (simply because they don't special-case much of anything), but otherwise we don't know of any current OS distributions that implement unicast for 127/8. The Linux kernel has long had a non-default "route_localnet" option offering similar but more problematic behavior. I would be happy to fund or run a project that would announce small global routes in each of these ranges, and do some network probing, to actually measure how well they work on the real Internet. We are working with RIPE Atlas already to enable this. I thought it would be prudent to propose the changes in detail to IETF, before "bogus" routes showed up in BGP and the screaming on the NANOG list started. :-/ John
Re: Redeploying most of 127/8, 0/8, 240/4 and *.0 as unicast
Fred Baker wrote: > My observation has been that people don't want to extend the life of > IPv4 per se; people want to keep using it for another very short time > interval and then blame someone else for the fact that the 32 bit > integers are a finite set. It's an attractive strawman, but knocking it down doesn't contribute to the discussion. I am not blaming anybody for the finity of 32-bit integers. Nor am I trying to extend the lifetime of IPv4, for either a short or a long time. Personally, I think that IPv4 will already be with us for the rest of my lifetime. Its life will already extend beyond mine, without any effort on my part. It was a good design, and it will outlive its makers. The people who in 2008 predicted that it was senseless to improve IPv4 because it would be dead by 2018, were objectively wrong, because it's not dead. IETF did what the objectors said, back in 2008: They didn't improve IPv4, on the exact theory that effort would go into IPv6 instead. Hmm. 13 years later, that decision did not cause IPv6 to take over the world and obsolete IPv4. IPv4 is still here, and the improvements that would have been fully rolled out to the user base by now, if standardized in 2008, are still missing. Perhaps we should reconsider that wrong advice, rather than regurgitate the same decision every time the issue is raised? > If you don't think that's a true statement, I'd be very interested to > hear what you think might be true. IPv6 is still on a remarkable if not meteoric growth ramp; in the last year it's gone from 30% to 34% of the Internet, according to Google. There is no need to cut off IPv4 maintenance at the knees in order to make IPv6 look good. We can make both v4 and v6 better, and let people choose which one(s) they prefer to use. That's what I think might be true about why simple low-risk tweaks that make IPv4 better are not an obviously stupid tactic. As Brian Carpenter has been saying for 15+ years, IPv4 and IPv6 don't have a transition strategy, they have a co-existence strategy. Neglecting or abandoning IPv4 is not a useful part of that strategy. Keeping the price of IPv4 addresses reasonable means that dual-stack servers can continue to be deployed at reasonable cost, so that it doesn't matter whether clients have IPv6 or IPv4. Any company that put its services on IPv6-only sites today would be cutting off 65% of their potential customers. Even if v6 had 90% of the market, why would a company want 10% of its prospects to be unable to reach its service? (Big companies who run massive public-facing server farms are the biggest buyers of IPv4 addresses in the market, spending hundreds of millions of dollars every year. Amazon, Google, Alibaba, Microsoft, etc. They are already running IPv6, and all their servers already have free IPv6 addresses from a RIR. Why are they spending that money? It isn't because they are stupid doodooheads who hate IPv6.) As Fred points out, this issue has been discussed since 2008. IETF also faced it in 2014-2018 in the now-sunsetted "sunset4" working group. That group wrote a bunch of drafts proposing to disable or sunset IPv4. None of these became RFCs. They didn't pass the sniff test. They weren't what users and customers wanted. The issue keeps coming back because the wrong decision keeps getting offered: "Just switch everybody to use IPv6, and if they won't, then deny their proposals for IPv4." Another way of putting it was proposed by Dave Thaler: "Rather than changing any existing software (OS's or apps) to allow class E addresses, the effort would be better spent towards getting more IPv6 deployment." This is not an objection to deploying reserved addresses in IPv4, it is a plea for more IPv6 deployment. It is like saying, "Do not spend your time fixing the environment, instead we need to fix the political system." It is a hopeless plea, since "allowing Class E addresses" and "more IPv6 deployment" are not the only two possible goals to put forth effort on. Merely stopping work on one, will not cause the other to be advanced. There must be a name for this fallacious argument...thank you, Wikipedia, it's a "false dilemma": https://en.wikipedia.org/wiki/False_dilemma The two goals can proceed in parallel, and many of the people who would happily do Goal 1 are not in any position to affect Goal 2 -- just as with the environment and politics. It's pretty simple to understand why IPv6 has not taken over the world yet. IPv4 got rapid adoption because it was so much better than all the alternatives available at the time. IPv6 would have gotten equally rapid adoption if it had been the thing so much better than all the alternatives. IPv6 is not getting that rapid adoption today, because it has to compete with the already pervasive IPv4. IPv4 is better in one key way: everybody you want to talk with is already there. (It's akin to the issue of why can't people just switch to a better social network than Facebook?
Re: Redeploying most of 127/8, 0/8, 240/4 and *.0 as unicast
Steven Bakker wrote: > The ask is to update every ip stack in the world (including validation, > equipment retirement, reconfiguration, etc)... This raises a great question. Is it even *doable*? What's the *risk*? What will it *cost* to upgrade every node on the Internet? And *how long* might it take? We succeeded in upgrading every end-node and every router in the Internet in the late '90s and early 2000's, when we deployed CIDR. It was doable. We know that because we did it! (And if we hadn't done it, the Internet would not have scaled to world scale.) So today if we decide that unicast use of the 268 million addresses in 240/4 is worth doing, we can upgrade every node. If we do, we might as well support unicast on the other 16 million addresses in 0/8, and the 16 million in 127/8, and the other about 16 million reserved for 4.2BSD's pre-standardized subnet broadcast address that nobody has used since 1985. And take a hard look at another hundred million addresses in the vast empty multicast space, that have never been assigned by IANA for anybody or anything. Adding the address blocks around the edges makes sense; you only have to upgrade everything once, but the 268 million addresses becomes closer to 400 million formerly wasted addresses. That would be worth half again as much to end users, compared to just doing 240/4! That may not be worth it to you. Or to your friends. But it would be useful to a lot of people -- hundreds of millions of people who you may never know. People who didn't get IP addresses when they were free, people outside the US and Europe, who will be able to buy and use them in 5 or 10 years, rather than leaving them unused and rotting on the vine. We already know that making these one-line patches is almost risk-free. 240/4 unicast support is in billions of nodes already, without trouble. Linux, Android, MacOS, iOS, and Solaris all started supporting unicast use of 240/4 in 2008! Most people -- even most people in NANOG -- didn't even notice. 0/8 unicast has been in Linux and Android kernels for multiple years, again with no problems. Unicast use of the lowest address in each subnet is now in Linux and NetBSD, recently (see the drafts for specifics). If anyone knows of security issues that we haven't addressed in the drafts, please tell us the details! There has been some arm-waving about a need to update firewalls, but most of these addresses have been usable as unicast on LANs and private networks for more than a decade, and nobody's reported any firewall vulnerabilities to CERT. Given the low risk, the natural way for these unicast extensions to roll out is to simply include them in new releases of the various operating systems and router OS's that implement the Internet protocols. It is already happening, we're just asking that the process be adopted universally, which is why we wrote Internet-Drafts for IETF. Microsoft Windows is the biggest laggard; they drop any packet whose destination OR SOURCE address is in 240/4. When standards said 240/4 was reserved for what might become future arcane (variable-length, anycast, 6to4, etc) addressing modes, that made sense. It doesn't make sense in 2021. IPv4 is stable and won't be inventing any new addressing modes. The future is here, and all it wants out of 240/4 is more unicast addresses. By following the normal OS upgrade path, the cost of upgrading is almost zero. People naturally upgrade their OS's every few years. They replace their server or laptop with a more capable one that has the latest OS. Laggards might take 5 or 10 years. Peoples' home WiFi routers break, or are upgraded to faster models, or they change ISPs and throw the old one out, every 3 to 5 years. A huge proportion of end-users get automatic over-the-net upgrades, via an infrastructure that had not yet been built for consumers during the CIDR transition. "Patch Tuesday" could put some or all of these extensions into billions of systems at scale, for a one-time fixed engineering and testing cost. We have tested major routers, and none so far require software updates to enable most of these addresses (except on the lowest address per subnet). At worst, the ISP would have to turn off or reconfigure a bogon filter with a config setting. Also, many "Martian address" bogon lists are centrally maintained (e.g. https://team-cymru.com/community-services/bogon-reference/ ) and can easily be updated. We have found no ASIC IP implementations that hardwire in assumptions about specific IP address ranges. If you know of any, please let us know, otherwise, let's let that strawman rest. Our drafts don't propose to choose between public and private use of the newly usable unicast addresses (so the prior subject line that said "unicast public" was incorrect). Since the kernel and router implementation is the same in either case, we're trying to get those fixed first. There will be plenty of years and plenty of forums (NANOG, IETF,
Redeploying most of 127/8, 0/8, 240/4 and *.0 as unicast
Steven Bakker wrote: > > ... the gain is 4 weeks of > > extra ip address space in terms of estimated consumption. > > The burn rate is the best argument I've seen against the idea so far. I'm glad you think so, since it's easy to refute. There will be no future free-for-all that burns through 300 million IPv4 addresses in 4 months. When IPv4 addresses were being given away for free, of course people were grabbing them as fast as they could. Particularly after everyone could see the end of the supply coming. It took detailed administrative bureacracy, checking paperwork "justifications", to just slow down the free-for-all! (In the early '90s I got 140.174/16 for the same cost as a /24, by sending a few emails asking for it. By the end of the '90s, that /16 was one of the major assets of our small ISP!) Now that has ended, and addresses actually cost money in a real market. Companies are only buying what they need, because they have other uses for their money. And other companies are selling addresses that they once obtained and no longer plan to use. It's like the difference between getting free land from the government, versus having to pay for it. You'd rather have 100 acres or 1000 acres if it's free. But if you have to buy it, well, half an acre is still pretty nice, and leaves you some money to build a house on it. Now we have a market. When low supply raises the price of something, people are going to buy less of it. The initial price rise from $0/each to $11.25 was in 2011, after ARIN announced there would be no more free addresses, and Microsoft bought Nortel's addresses out of bankruptcy. The Internet world has adapted. It's been a decade since that adaptation. Adding some global unicast address supply in a few years would reduce prices, benefiting consumers, but won't take us back to the old pre-market model. Here's an analysis as of late 2020 of the IPv4 transfer market: https://circleid.com/posts/20201125-ipv4-market-and-ipv6-deployment It shows a range of 6 million to 16 million addresses transferred (sold) per quarter. This roughly matches Geoff Huston's analysis of both free RIR allocations, and purchased IPv4 transfers. Geoff reports about 2.2 million allocated and 26 million transferred in 2020: https://circleid.com/posts/20210117-a-look-back-at-the-world-of-ip-addressing-in-2020 At current rates, 300 to 400 million addresses would last more than a decade! (Compare this to the ~50 /8s unadvertised in global routing tables, about 838 million addresses, that are currently owned but likely to be sold to someone who'll route them, sometime in the next decade.) There will be no future free-for-all that burns through 300 million IPv4 addresses in 4 months. John