Next up at #netdevconf: Low Latency Low Loss with scalable throughput.
Sounds too good to be true, but it's an experiment, let's see how it goes. :)
Improvements to UDP message receive calls presented at #netdevconf:
rcvmsg - Receive one message on one socket
recvmmsg - Receive multiple messages on one socket
--> recvmmmsg - Receive multiple messages on multiple sockets
I appreciate the naming :)
At #netdevconf there was also a talk proposing to make some more IPv4 address space usable:
- 240/4 and 0/8 which are "Reserved for Future Use"
- 127/8 and 224/4 which are under-utilized
- Zeroth address of every subnet and broadcast address of some subnets
They have running code, now they want to do the politics.
The #netdevconf keynote was given by Alissa Cooper, #IETF chair: "Open Source, the IETF, and You".
She told the story of joint development of standards and code for TLS 1.3 and QUIC, and then invited people in the Linux community to come and contribute in the IETF - joining the hackathon, writing drafts, or otherwise.
Now presented at #netdevconf by Hajime Tazaki: Is reimplementing the network stack a good idea? - Of course not!
With some protocol conformance measurements to back up the claim. Sadly using a closed-source tool.
Also, the protocol implementation conformance tests in this #netdevconf talk treated MAY and SHOULD as a MUST (i.e., counted fails for them), and I think the IPv6 RFCs on the slide were the obsoleted ones.
Nevertheless - reinventing the wheel is not a good idea.
IPv4 address space, netdev talk
The #netdevconf talk on IPv4 unicast address space expansion started with a motivation about the IPv4 address market and speculation.
Actually, adding more IPv4 addresses might make that problem worse, another participant (Mirja) just said. People are buying IPv4 addresses because they think they need them.
Sounds plausible to me.
@NerdResa I understand the sentiment, and I appreciate people doing such work. But we had this discussion in the numbers community trillions of times and it just pushes the problem away and reduces the pressure for people to deal with the actual problem (and yes they rather want horrible circumvention techniques ..cgn ftw 🙈 than rolling out v6). Plus this address spaces are "tainted" it's gonna be a mess getting these routed globally.... I wouldnt want to go there (my very personal oppinion)
@mutax Yea, I think this was one of the more difficult/controversial ideas. I don't expect all of these to work either.
I don't think there's even a draft yet though, so it won't be discussed this IETF I guess.
I think, if all the time and energy put into this went into deploying IPv6, we would get just as far.