Hacker News Reader

Comment by BlueTemplar | Hacker News Reader

So, why haven't the relevant people started working on IPv7/NewProtocol, like... 10 years ago, when it started to become clear that mobile Internet was going to become big and that IPv4/6 and TCP/UDP coudln't work well with it?

View on HN

These things take a very, very long time. Consider that IPv6 has been a thing since the late 90's. It is now older than the IPv4 internet was when it started taking off around 1995. IPv6 is it. I very much doubt we will see an IPv7 in our life times, if ever.

From my experience, it's because some of the people that would need to be doing that have their heads in the sand about e.g. the necessity of middleboxes and application-layer proxies. Network-layer folks still seem to think that the Internet must not have application-layer proxies, and some actively resist good efforts to expand the application-layer protocols to better deal with the reality of application-layer middleboxes, such as those widely used in IoT applications. And some try to eliminate good mechanisms that otherwise would work, such as CONNECT over TLS.

Probably because they noted the terrible time they were having getting people to migrate from IPv4 to IPv6, and had no desire to repeat it.

It'll only take another 50 years :')

They did! https://www.schneier.com/blog/archives/2010/06/darpa_researc...

The problem is that the economic situation makes it very difficult to replace the Internet. Wholesale protocol replacements only work if they are backwards compatible and incrementally deployable and they provide new capabilities even when there is partial deployment.

With most of the population buying a new phone every two years, 5G, and the carriers being able to be a compatibility layer, it seems like it wouldnt be hard at all to do something like QUiC/IP7 and have massive adoption within a couple years. Apple, Microsoft, Samsung, Google, ATT, and Verizon could force it on their own.

I think it was obvious in the mid to late 90's that ipv6 was flawed once the internet for civilians really started to take off, not having a seamless migration path was not on.

A seamless migration path (in the sense that you seem to be wanting it) was impossible. v4's design doesn't allow for it.

And your basing this on?

My knowledge of v4's design. The header fields for source and destination address are 32 bits wide, and the pigeonhole principle stops you from fitting more than 32 bits of addresses into those fields.

Reply on news.ycombinator.com