Comment: IPv6 obsolete? I think I'm standing in the forest!
IPv4 addresses are often only available at horrendous prices. Those who propose CDNs instead are leaving the network to the hyperscalers, says Martin Loschwitz.
(Image: iX)
- Martin Gerhard Loschwitz
The law on router freedom came into force on August 1, 2016. A Monday that I remember well: because Tempelhofer Hafen in Berlin was celebrating its birthday the day before, the MediaMarkt store there was open and I bought a Fritzbox 6490 for my cable connection at home. I had been preparing almost meticulously for weeks. For example, I was in possession of the activation code that Vodafone required at the time in order to actually be able to use my own modem on my home connection.
And I had a goal: anyone using the provider's own cable modem had the choice between DS-Lite without a real IPv4 address or IPv4-only access. Unfortunately, according to the provider, dual-stack with real IPv4 and real IPv6 is not technically feasible – unless you don't use the provider's modem but use your own cable modem. Nevertheless, the dual-stack setup at home did not work the next day. This was because the Fritzbox firmware at the time did not offer the option of activating an IPv6 address. After a brief period of mutual recriminations between Vodafone and AVM, the manufacturer delivered the feature a few weeks later and the dual-stack setup finally worked as desired.
Almost eight years later, the situation with regard to IPv6 has still not improved significantly in many places. The much-maligned Telekom now reliably delivers true dual-stack on many DSL and fiber optic connections. With other providers, however, annoyed customers continue to struggle with problems and adversities such as those described above. Many providers still don't care about IPv6, as I found out when I moved into an apartment in Berlin-Schöneberg in 2021: despite having a cable connection, you couldn't get a public IP address at all, neither v6 nor v4.
Instead, you had to deal with CGNAT of the worst kind, and even booking a fixed IPv4 didn't help: You could then be reached via IPv4, but only through a confusing DNAT construct on the provider side. If you use your Internet connection – in this case even a business customer connection – for more than Netflix and occasional surfing, this kind of connection junk simply doesn't help. Clemens Schrimpe sends his regards: connections of this kind are functionally broken.
Told you so!
Fittingly, the main person responsible for science at APNIC caused a stir last week with the thesis that he considers the goal of completely replacing IPv4 with IPv6 to be obsolete and unenforceable. The text quickly made the rounds within the Internet community. Not least because it shakes up what the networkers in our industry in particular have been preaching for years: Namely, that the complete switch to IPv6 was a sheer necessity.
In the hustle and bustle surrounding the announcement, however, it was overlooked that Greg Huston is by no means calling for IPv6 to be abandoned altogether. Even Huston's proposal for a new goal, that IPv6-only ISPs should be able to exist and be useful, is still miles away. It will hardly help to reduce the targets that large parts of the industry have been stubbornly ignoring for years anyway. However, statements such as those made by Greg Huston are likely to serve as a new basis for the arguments of IPv6 deniers. They have always known that IPv6 would not catch on. Huston is thus doing a disservice to those who consistently implement and use IPv6.
Especially as Huston's argumentation inadmissibly abbreviates facts in many places and generously ignores the problems that would result from the proposed action. Of course, from the perspective of an ARIN or RIPE NCC member, it is easy to sit on one's lazy ass and simply ignore the IPv4 problem. This is because both registrars are generously equipped with IPv4 blocks in which there are still some reserves lying dormant.
Elsewhere in the world, the IPv4 situation is much less relaxed. APNIC in particular, which is responsible for the Asia-Pacific region, has repeatedly complained in the past that the number of IPv4 addresses theoretically available per inhabitant there is very low. This is because APNIC is a priori only equipped with a few, relatively small address blocks. They are therefore aware of the problem, which makes Greg Huston's statements all the more incomprehensible. This is one of the reasons why IPv6 is proportionally much more widespread in the APNIC area today than in the ARIN or RIPE catchment area: out of sheer necessity.
And even here in RIPE-land, the IPv4 situation is no longer as relaxed as it once was. A lively trade in IP networks has long since established itself – and the prices are steep: even ASN numbers, which entitle you to participate in the network as an "autonomous system" in the first place, now go for thousands of US dollars over the counter. An IPv4 network from RIPE costs 32 euros per IP, whereby only larger networks such as those with 1024 IP addresses (/22) are usually sold. The fun must therefore be worth just under 33,000 euros, while IPv6 networks are negligible in terms of price but offer the same functionality. The horrendous costs for IPv4 networks can hardly be afforded by smaller market players in particular. Ultimately, the lack of IPv4 even jeopardizes net neutrality, because under these circumstances not everyone has or can have the same access to the network.
Videos by heise
Internet, not CDN-Net
Greg Huston's statement that the relevance of connectivity on the Internet as a whole is declining and that it would be better to speak of a "network of services" in future is highly dangerous for the same reasons. If we come to terms with the fact that the majority of content on the Internet will only reach users via hyperscalers or large CDNs, these providers will at some point feel called upon to set the rules of the game themselves. Especially as these institutions also have the war chest to absorb IPv4 capacities that become available and simply cut off the path to potential customers for alternative or new providers. In practical terms, this is therefore nothing less than the introduction of an IT oligopoly through the back door. And sorry, but is that still possible?
The alarm bells should have been ringing for a long time (not only) at European institutions and authorities if the digital sovereignty that they themselves have promised so loudly was taken seriously and actually sought to be achieved. After all, nobody can want managers in back rooms in San Francisco, Mountain View, Redmond, Seattle or anywhere else to decide what content is delivered to whom anywhere in the world and how. But that is exactly what will happen as soon as the respective providers have the technical ability to do so. Incidentally, this has already happened in the USA at least, as AOL essentially amounted to this in the pioneering years of the Internet – with the well-known result. As things stand, IPv4 not only offers a relatively easy way to make a more radical attempt at a new edition, but also a way with no alternative for end users.
Instead of looking for reasons not to introduce IPv6 after all, providers and, not least, legislators should get their act together and push ahead with IPv6. There is no alternative if we want to preserve and further develop the Internet in its current form instead of turning it into a cesspit along the lines of Twitter. A lack of IPv6 connectivity must be considered a shortcoming of an Internet connection and it must finally be more expensive for ISPs to ignore IPv6 than to make an effort to integrate it sensibly. Only in this way will it be possible to leave a global network to future generations whose versatility and practical benefits match those of the present.
(vbr)
