Conversation

Notices

1. Embed this notice
   famfo (famfo@chaos.social)'s status on Monday, 24-Mar-2025 03:15:16 JST famfo

   I'm not even mad at CGNAT anymore, I'm just really mad at the IPv6 adoption and missing resources around it.

   For most of my friends who did not take the deep dive into computers and especially computer networks, an IP address is four octets containing funny numbers.

   A friend of mine got really confused when I mentioned that there is another IP version: IPv6 (which of course did NOT work for the use case we were trying to achieve: making proprietary software talk with eachother).

   • Haelwenn /элвэн/ :triskell: and Doughnut Lollipop 【記録係】:blobfoxgooglymlem: like this.
   • Embed this notice
     fossdd @ FOSDEM (fossdd@chaos.social)'s status on Monday, 24-Mar-2025 03:16:00 JST fossdd @ FOSDEM

     in reply to

     @famfo this is so fucking sad. like ipv6 has been around for almost forever, but in most usecases its still considered second-tier. heck even in the current curriculum of my abitur we still learn mainly classfull!!! ipv4 ("obsolete network addressing architecture used in the Internet from 1981 until the introduction of CIDR in 1993"). how is this possible? im writing my exams in two months and i wont get points if i assume classless networking??

     we're fucking stuck in 1993

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Monday, 24-Mar-2025 03:20:56 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • fossdd @ FOSDEM
     @fossdd @famfo Reminds me of how often people refer to OSI layers… even though OSI was dead on arrival and TCP/IP was the one that actually worked.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Tuesday, 25-Mar-2025 23:39:13 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • muxelplexer
     • fossdd @ FOSDEM
     @muxelplexer @famfo @fossdd Except there's no such thing as layer 1/2/…/5/6 on TCP/IP.
     
     You can put whatever you want in-between when it comes to TCP/IP, in any-amount, after all HTTPS didn't need a protocol change for HTTP.
     
     And it's not just in TCP payloads, which layer exactly is PPPoE? :D
   • Embed this notice
     muxelplexer (muxelplexer@larkspur.one)'s status on Tuesday, 25-Mar-2025 23:39:14 JST muxelplexer

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • fossdd @ FOSDEM
     @lanodan @famfo @fossdd but isnt the issue with TCP/IP the combination of OSI layer 1 & 2? Not like anyone will care about OSI layer 5/6 these days either but we've been told thats why both are still around battling around
   • Embed this notice
     翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Tuesday, 25-Mar-2025 23:44:36 JST 翠星石

     in reply to

     • fossdd @ FOSDEM
     @fossdd @famfo You need to be actually somewhat competent to understand IPv6 and it seems many people are so incompetent they go disable IPv6 and use IPv4 only.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Wednesday, 26-Mar-2025 00:04:45 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • muxelplexer
     • fossdd @ FOSDEM
     @muxelplexer @famfo @fossdd I do think there's a bit of value in doing the difference between physical and internet, but you don't need OSI for this.
     OSI can be left out with X.25 and the other protocols it had at the time.
   • Embed this notice
     muxelplexer (muxelplexer@larkspur.one)'s status on Wednesday, 26-Mar-2025 00:04:46 JST muxelplexer

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • muxelplexer
     • fossdd @ FOSDEM

     @lanodan @famfo @fossdd generally you can just use either and make the best of both worlds by acknowledging the OSI split between physical/link and the TCP/IP amalgation of application layer - but in terms of naming the layers imo OSI is superior

   • Embed this notice
     muxelplexer (muxelplexer@larkspur.one)'s status on Wednesday, 26-Mar-2025 00:04:47 JST muxelplexer

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • fossdd @ FOSDEM

     @lanodan @famfo @fossdd

     Except there’s no such thing as layer 1/2/…/5/6 on TCP/IP. Yeah - they aren’t numbered. There is still Link/Internet/Transport/Application Layers though. Call them what you want but imo that’s a bit pedantic lol.

     Yes - but that only applies from the Transport layer upwards - as it does with OSI. In OSI we just ignore Session/Presentation since they are obsoleted. Once you go downwards you have abstracted a physical electrical connection together with a link-layer MAC connection - which is not beneficial when working on those lower layers.

     PPPoE is encapsulation of Layer 2 frames in Layer 3 packets in OSI terms, which is nothing too mind bending.

     ((sorry if i come off as a bit annoyed but OSI layer model is more than just a reference for programmers and i may have PTSD on the layer systems from Networking classes :p ))

   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Wednesday, 26-Mar-2025 00:08:49 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • muxelplexer
     • fossdd @ FOSDEM
     @muxelplexer @famfo @fossdd And I think trying to wrangle the TCP/IP and somewhat related protocols into OSI layers (be them numbers or names, doesn't matters) is trying to make round pegs fit into a square hole.
   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Wednesday, 26-Mar-2025 00:16:41 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • muxelplexer
     • fossdd @ FOSDEM
     @muxelplexer @famfo @fossdd Not good at pre-RFC history of the internet but as far as I can tell DARPA was even older than that (I know there was UUCP and NCP).
     
     And OSI (which is from ISO) would mean new protocols, and as far I can tell, it failed to actually get deployed outside of labs.
     
     While TCP/IP managed to just encapsulate older protocols (as seen by it supporting the dinosaur that is FTP).
   • Embed this notice
     muxelplexer (muxelplexer@larkspur.one)'s status on Wednesday, 26-Mar-2025 00:16:42 JST muxelplexer

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • muxelplexer
     • fossdd @ FOSDEM

     @lanodan @famfo @fossdd also on a side note: wasn’t it technically the DARPA networking model instead of TCP/IP or do i have brainworms?

   • Embed this notice
     muxelplexer (muxelplexer@larkspur.one)'s status on Wednesday, 26-Mar-2025 00:16:43 JST muxelplexer

     in reply to

     • Haelwenn /элвэн/ :triskell:
     • fossdd @ FOSDEM

     @lanodan @famfo @fossdd i can get that sentinment - it probably feels natural to me since that was how our networking prof got us to layer the networking (coming from a infra/administration background). Ignoring Layer 2 is just not possible when you operate with non-multi-layer switches (i.e. good ol switches)

   • Embed this notice
     翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Wednesday, 26-Mar-2025 13:26:22 JST 翠星石

     in reply to
     @famfo There are differences, there is diffences in subnetting and different configuration required.
     
     With IPv4, most stuff is already automatically configured for you, thus you usually don't need to do much to configure an IPv4 router, but for some bizarre reason, IPv6 routers need manual configuration usually.
   • Embed this notice
     famfo (famfo@chaos.social)'s status on Wednesday, 26-Mar-2025 13:26:24 JST famfo

     in reply to

     • 翠星石

     @Suiseiseki the only difference between IPv4 and IPv6 is the representation of the bits and that one is 32 bit and the other 128.

     In no way does someone need to be "more competent" to use IPv6, I am fully convinced that it is just a lack of resources around it, be it educational, transferal from existing IPv4 knowledge or the general rollout on bigger websites.

   • Embed this notice
     翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Wednesday, 26-Mar-2025 15:19:06 JST 翠星石

     in reply to

     • gentoobro
     @gentoobro @famfo If you're typing in a IPv6 address in (that's not :: or ::1), you're doing something wrong.
     
     There's something called domains or GNUnet.
     
     getaddrinfo(), connect(), bind() and listen() should have been modified to support IPv6 without needing specific configuration, or at least be compatible if you just change AF_INET to AF_INET6, but noooo.
   • Embed this notice
     gentoobro (gentoobro@shitpost.cloud)'s status on Wednesday, 26-Mar-2025 15:19:08 JST gentoobro

     in reply to

     • 翠星石

     Most software has to be modified specifically to handle IPv6. There is some stupid wrapper shit that kinda works a little bit sometimes for some apps, but you can't type an IPv6 address into software that only supports IPv4.

   • Embed this notice
     翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Thursday, 27-Mar-2025 13:12:33 JST 翠星石

     in reply to

     • LisPi
     • gentoobro
     @lispi314 @famfo @gentoobro Which is why I mentioned GNUnet domains.
   • Embed this notice
     LisPi (lispi314@udongein.xyz)'s status on Thursday, 27-Mar-2025 13:12:34 JST LisPi

     in reply to

     • 翠星石
     • gentoobro
     @Suiseiseki @gentoobro @famfo Domains introduce an authority problem in their most common form.
     
     They also introduce authentication problems as it is mostly not used in conjunction with key-addressing.
     
     Effectively, user-configured petnames (ref: Zooko's Triangle) and key addressing as Yggdrassil does is the only sane way to communicate at the application level (yes the implication is that there's no sane way to use IPv4 as it cannot encode a sufficient key).
     
     The keys need not be encoded in IPv6, the underlying transport layer may be abstracted. But applications should not concern themselves with routing details (unless they are specifically overlay network programs, of course).