1y ago

Happy 30th birthday to RFC 1631 ("NAT"), the "short term solution" we all rely on

datatracker.ietf.org

RFC 1631: The IP Network Address Translator (NAT)

From the conclusion:

NAT may be a good short term solution to the address depletion and scaling problems. This is because it requires very few changes and can be installed incrementally. NAT has several negative characteristics that make it inappropriate as a long term solution, and may make it inappropriate even as a short term solution. Only implementation and experimentation will determine its appropriateness.

32 comments

there is no fix more permanent than a temporary one.
edit: as I literally sit here inspecting the nat tables on a couple of edge routers.
- That temporary fix will eventually become unnecessary. IPv6 has slowly getting more and more use.
  
  This thread starts with a document literally proving people have been saying that exact thing for 30 years now.
  
  It's been getting "more and more use" since 2001. To start with the isps said that they were not going to do any work to implement it until endpoints supported it. Then vista came with support by default. Next they wanted the backbones to support it. All tier 1 networks are now dual stack. Then they said they were not going to do anything until websites supported it widely. Now all cdns support it. Then they said, it's ok we will just do mass nat on everyone so won't do any work on it.
  
  Very, very slowly.
  
  So has Linux on the “desktop” buts it’s never been the year of the Linux desktop.
- My previous office was in a set of partitions put up in a library 20 years ago as a temporary measure.
I wrote so many essays and exam answers in the late 90s on how IPv6 would come in and fix everything and I’m really feeling this.
- It did, wherever it's used. If you can ditch backwards compatibility in your network and just use ipv6, everything gets so much simpler.
192.168.1.1/24. Got it.
Everyone
- I've only recently branched out from router defaults...only reason was that I wanted to VLAN off my home network, and mostly just so [Home Assistant-controlled] smart devices can't talk to the Internet at all.
  
  10.0.0.0/8
  172.16.0.0/12
  192.168.0.0/16
  🎶 a whole new wooorrrld... 🎶
Ah, how to forget the first obstacle in my hobby self hosting projects, the damn CGNAT....
"Just open the wireguard port bruh"
No my friend, I don't think that is gonna cut it.
(Thankfully Zerotier and Tailscale work for me).
- I have the same issue (TRIPLE NAT'd! One of which is the CGNAT). Unfortunately I have external family that accesses from media boxes/TVs so those won't work for me.
  Thankfully I was able to get a small VPS server for $2/mo and set up some reverse tunnels with auto-ssh. Seems to be working fairly well so far.
  All that said, I longingly look forward to the future when I don't have to worry about NAT.
  
  What's really crappy is that my ISP which used to give me a public ipv4 and also supported ipv6 2as bought out, and now I'm on cgnat and ipv6 support has disappeared.
  Fuck metronet, it's not even cheap anymore
A few people need to get off their horses and come up with and agree to IPv4². It's exactly the same as IPv4 except there's 2 more octets of address space - 48bits for addresses*. Job done. You'd see wide spread adoption in under 2 years and then we can forget about it all and move on with our lives safe from the clutches of IPv6.
I don't give a crap that doesn't neatly fit into 32 or 64 bit architectures. It's more than doable at plenty fast speed and it keeps everything manageable.
- And what would be the advantage? It wouldn't be routable through legacy systems, and you'd run out of addresses in a couple of years again.
- What's bad about IPv6?
  
  Nothing. It fixes the myriad of horrible hacks that are required for ipv4 to somehow still hang on.
  Of course companies are sad because transition costs money, even though as usual the open source community did most of the work for them.
  
  https://youtu.be/oSIPiB2JEkA
  
  It tried to fix everything wrong with IPv4, like shitty multicasting. This made it extra complicated.
  If it had just been 128-bit addresses, it probably would have been widely deployed in the 90s. Don't need to bother at this point, though, just get it done.
- If we rolled back the clock to 1998 or so, that'd probably be worthwhile. At this point, everything is set for IPv6, and we just need to do it.
A day old IPv6 thread where there isn't some moron arguing NAT is for security? What's going on here?
Thing is I knew it as masquerade for years before I heard the term nat.
- Linux IPchains from the 2.2 kernel days?
  
  Early version of slackware from around 95 or so.

32 comments