@jschauma @ryanc @0xabad1dea Back in 2018 Dave Taht and a few other people tried to reclaim the "class E experimental" address range and most of multicast, which together are about 1/8 of the IPv4 address space.

https://landley.net/notes-2018.html#25-10-2018

The ipv6 stans FREAKED OUT because the only argument their terrible tech ever had was ipv4 exhaustion, largely due to poor initial allocation back in the 1980s (giving korea 64k total addresses, for example).

@landley @jschauma @ryanc @0xabad1dea I think #IPv6 would've gotten more acceptance if it was merely a 4x long #IPv4 annotation instead of doing hexadecimals.

• I mean, worse would've only been #base32hex or #base64hex for IPv6 addresses...

@kkarhan @jschauma @ryanc @0xabad1dea Nope, ipv6 is fundamentally flawed because you can't persistently identify an internet access point in a useful way. That's why wikipedia blocked the entire IPv6 address range for anonymous edits. (May still do, haven't checked.)

You could instead have subdivided the port space without ANY protocol change, and done 1.2.3.4.[0-16] to give each NAT user their own 4096 public ports. 1.3.2.4.2 port 80 is a web server on host port 8192+80...

@kkarhan @jschauma @ryanc @0xabad1dea Ahem, [0-15].

As the saying goes: the two fundamental problems in computer science are cache invalidation, naming things, and off by one errors.

@kkarhan @jschauma @ryanc @0xabad1dea There's only about twice as many people on the planet as there are IPv4 addresses NOW. Increasing the address space by 16x would mean each person (including infants and the illiterate) could have 8 public devices online fulltime without sharing or dynamic provisioning.

Giving each grain of sand in the solar system its own subnet was unmanageable futurism BS to extend the address range to a star trek future with zillions of planets talking via instant FTL.

@pmevzek @landley @jschauma @ryanc @0xabad1dea still, a #64bit #AddressSpace would've been more than sufficient as we can see by the fact that /64 is the default #IPv6 allocation for basically any consumer connection.

A #128bit address space is quite inefficient given we ain't saturating even half of it.

#ZFS does have that problem, abeit #Sun engineers at the time expected #64bit to be as quickly deprecared as #16bin and #32bit.

@pmevzek @kkarhan @jschauma @ryanc @0xabad1dea Some friends in Tokyo talked to a hosting place at the top of a skyscraper in shinjiku to get 3U of rack space last week, and part of the negotiation was how many IPv4 addresses they should get for the servers. (They were offered up to 8 depending on what they wanted to pay.)

I asked: the japanese translated in their .en PDF as "publicly routable" meant ipv4. IPv6 isn't even used internally there (it's 192.168.x.x or something behind the firewall).

@landley @pmevzek @jschauma @ryanc @0xabad1dea And that is espechally baffling considering that #APNIC as #RIR ran out of #IPv4|s first.

• And right now I'm between a rock and a hard place and either become a @ripencc member or cough up equal amounts in monthly installments just to get some properly geolocated #IPv6.

Cuz people telling me "Oh, just use HEnet's Tunnelbroker"" never experienced the shitshow that is #Geoblocking and #GDPR-#Noncompliance with #Malvertising to slap their faces!

• Cuz believe me, I tried, but since some idiots decided to #GeoIP entire #ASN|s and not #IP - #Allocations the PoP in FRA (FFM actually) will get me mislocated to the #USA!

@kkarhan @pmevzek @jschauma @ryanc @0xabad1dea @ripencc j-core.org is 5.8.71.57, and landley.net is 67.205.27.143. Neither had trouble getting or keeping an ipv4 address for many consecutive years. There's over a billion of them, even with IPv6 advocates keeping hundreds of millions of multicast and class e addresses out of use.

We haven't suddenly needed to quadruple the length of phone numbers or credit card numbers either. That's not how humans work.