Unallocated Class E IPv4 Addresses
by Leo Vegoda
The internet was relatively new in the mid-1980s, and still largely dominated by academic institutions. Its future growth was – literally – unimagined. The people developing the protocols we continue to use today didn’t know how the internet and its support pieces would evolve. They just knew that they needed to be good stewards.
The 268 million unallocated Class E IPv4 addresses – 16 /8 blocks – are just one of the results of that response. That’s 268,777,216 addresses.
Multicast
IP multicast is a protocol that lets one data packet go to many receivers. A benefit is that it can lower bandwidth requirements. When describing multicast in 1986, its designers also reserved a big block of addresses for “future addressing modes”. After all, the designers were aware of their inability to predict the future. They called this reserved space, Class E.
IP multicast is still used on financial trading platforms and for video distribution. But those future good ideas that could have used Class E space were never developed. So, addresses in the range 240.0.0.0 to 255.255.255.255 have been reserved for almost 30 years.
Exhaustion
In 2008 people were worrying about the rapid pace of IPv4 allocation. The global IPv4 address reservoir was about to run dry, leaving the Regional Internet Registries (RIRs) to run down their stock. In response to pending exhaustion of IPv4 supply, teams published two slightly different proposals for how Class E IPv4 addresses could help with the threat.
The more conservative proposal came from an APNIC team. They proposed to formally scope Class E addresses as additional private space. That meant it could be used on private networks but not routed across the internet.
The more radical proposal came from a team at Cisco. They proposed to redesignate the space as unicast. Unicast means one-to-one communication, while multicast is one-to-many. Unicast is what we use when we browse the web, stream music, or send a message.
The Cisco proposal’s key difference was that it did not propose a scope. Instead, it invited a discussion on potentially designating Class E addresses as globally scoped. That would mean allocating them to RIRs, so they could be used in homes, offices, and data centers around the world. They “envisioned that the utility of this block will grow over time.” They also noted that “some devices may never be able to use it.”
The devices that might never be able to use addresses from the Class E space were deployed across the internet and not managed or controlled in a centralized way. They included computers whose operating systems would need simple updates to turn on support for these addresses. But they also included devices that often don’t get updates, like home Wi-Fi routers.
The existence of those devices, along with IPv6 as an alternative, have blocked both of these proposals. Why waste time on updating systems to cope with just 16 /8s when there’s so much IPv6 address space available?
Because the IPv4 Class E space was not redesignated, the last five IPv4 unicast /8s were allocated to the RIRs in February 2011. Neither the APNIC nor the Cisco proposal progressed and the Class E space is still formally reserved. A 2011 document noted that “it is possible that 240.0.0.0/4 might only be useful in very large, new greenfield deployments where full control of all deployed systems is available.” Which is to say, that the IPv4 Class E addresses would be too unreliable for internet use. Instead, they were best suited to use in an environment controlled by a single organization.
The Cloud
Cloud computing has a history going back to the 1960s and got that name in the 1990s. But it was a specialist service until the early 2000s, when Amazon AWS and Google Docs hit the market. They h it the description of the greenfield deployment envisaged in the 2011 document.
But use of the reserved Class E space was put on hold pending decisions about its future use. While it is formally reserved, it’s not technically impossible to use. It was there and so got used. In 2022, internet engineers detected use of Class E addresses in private networks (not connected directly to the internet) run by Amazon AWS and Adobe. In 2024, an APNIC team ran an experiment to see whether internet end users could access services hosted in Class E space. So, the space was reserved but could be quietly used in private networks.
APNIC’s 2024 experiment showed that just 0.0452 percent of users could access their Class E test download. This is based on a sample size of 130 million users. They concluded that “the status quo is entirely adequate for the 240/4 address prefix!”
Of course, most networks won’t send data to these addresses and will reject packets coming from them. That’s because they are formally reserved. Allowing access to reserved addresses could put users at risk of malware downloads and worse. Preventing propagation of incorrect routing information is so important that it’s the first action on the MANRS requirements list. MANRS is a global initiative to reduce the most common threats in internet routing.
So, how would things be different if these addresses were allocated to RIRs? There would be two main issues. Firstly, creating policies for allocating these addresses would not be simple. Secondly, making them usable would be a challenge.
The current policy for allocating IPv4 addresses to the RIRs assumes a small and decreasing pool of reclaimed addresses. And that policy is struck through on the ASO website because the pool is empty.
Allocating Class E
Nonetheless, some people desperately want more IPv4 addresses. 268 million new addresses looks like an appealing prospect if you don’t look too closely. But how could they be put into the hands of the networks that want to use them on the internet?
If 16 new /8s became available then a new policy would be needed to allocate them. The process for developing this kind of Global Policy generally takes two years.
16 of anything, including /8s, cannot be easily broken up into five equal pieces, so giving each RIR the same amount won’t work. The previous Global Policy used an allocation trend calculation to determine how many /8s an RIR should get. But the IPv4 market has been actively redistributing IPv4 addresses for the last decade, so allocation trends wouldn’t make sense.
The RIRs themselves have policies to evaluate how many IPv4 addresses a requester can justify. Essentially, requesters of more space need to demonstrate a need and the ability to usefully deploy the new IPs. While RIR justification rules vary, they are all similar to guidelines documented in 1996. Where they have been removed from policy documents, they could quickly be reinserted.
But managing a flurry of applications for a limited resource would be an administrative challenge. The IPv4 runout experience suggests that there would be a rapid influx of membership requests, initial allocation requests, and mergers. While many requests would doubtless come from organizations that want to run networks, others would come from companies seeking to stockpile and resell IPv4 addresses.
This was the experience when RIPE NCC made any new entity eligible for a /24 simply by requesting it. Given the value of this block it made great sense for the entrepreneurial to establish shell entities, take possession of the /24 blocks and wait the required two years to then sell the IPs.
Developing and implementing policies doesn’t solve the second problem: making the addresses usable. Internet systems have not had to support Class E addresses over the last 30 years. Some systems are never updated, only replaced. There’s a problem when those systems sit between people using Class E addresses and people who want to connect to them.
That doesn’t mean that most business and consumer operating systems would reject them. Most of their code has been updated. But the machines in the middle generally get fewer updates. They are infrastructure, and so stability is valued. Many networks only apply updates to fix bugs affecting them. And lots of home and small office networking equipment is effectively unmanaged.
The people whose systems are numbered with Class E addresses would have to convince others to spend money on fixes. Essentially, they’d be asking organizations running incompatible systems to obtain, test, and install an update. Or, in many cases, replace equipment for which updates aren’t available. The costly updates wouldn’t benefit those spending the money. Experience has shown that this approach is a “challenge.”
Meanwhile, the large organizations already using these addresses in a private context would need to make changes, too. Use by Amazon AWS and Adobe has been detected but that doesn’t mean they are alone. Or they might choose not to make those changes, calling the bluff of other networks.
Their kind of unofficial use has been seen and measured before. But there has been considerable market consolidation since then and the internet is not in a rapid growth phase. It has become essential infrastructure.
So, if this space were allocated, the competition authorities empowered by governments could find themselves choosing between two options. One would be to favor new market entrants. To do so would mean to support them getting some IPv4 addresses at bargain prices. The other would be favoring large enterprises – and their customers – with significant investments in the unofficial use already in place.
Of course, the organizations using Class E space already would participate in discussions about redesignating it. Their engineers can argue against it and could be persuasive. The cost of making these addresses reliably useful for anything more than private use would be unpredictably expensive. And the benefits would not be certain either.
Formally designating them as additional private addresses does no harm. But as they are being used like that already, it would just be a paperwork exercise.
Change is unlikely and impractical. IPv6 sees a growing share of internet traffic, averaging 40 percent at the moment. That’s the place to invest effort.