I've recently learned about jumbograms (UDP packets > 65k - IPv6 only). Although I have no requirement to use them I thought I'd dig around and see how people had done anything clever.
I was surprised to find very little other than technical documentation protocol.
So has anyone used them and if so how?
Are they less reliable than UDP because they're larger?
Which language / software / hardware supported them?
What was your experience?
Cheers
Linux includes support as there is a vulnerability logged against it:
Systems affected
Linux Linux Kernel: 2.6.21, Linux Linux Kernel: 2.6.20, Linux Linux
Kernel: 2.6.2, Linux Kernel: 2.6.2 rc3, Linux Kernel: 2.6.2 rc2, Linux
Kernel: 2.6.2 rc1, Linux Linux Kernel: 2.6.20.1, Linux Linux Kernel:
2.6.20.10, Linux Linux Kernel: 2.6.20.11, Linux Linux Kernel: 2.6.20.12, Linux Linux Kernel: 2.6.20.13, Linux Linux Kernel: 2.6.20.14, Linux Linux Kernel: 2.6.20.2, Linux Linux Kernel: 2.6.20.3, Linux Linux Kernel: 2.6.20.4, Linux Linux Kernel: 2.6.20.5, Linux
Linux Kernel: 2.6.20.8, Linux Linux Kernel: 2.6.20.9, Linux Linux
Kernel: 2.6.21.1, Linux Kernel: 2.6.20 rc2, Linux Linux Kernel:
2.6.20.15, Linux Linux Kernel: 2.6.20.6, Linux Linux Kernel: 2.6.20.7, Linux Kernel: 2.6.21 git1, Linux Kernel: 2.6.21 git2, Linux Kernel:
2.6.21 git3, Linux Kernel: 2.6.21 git4, Linux Kernel: 2.6.21 git5, Linux Kernel: 2.6.21 git6, Linux Kernel: 2.6.21 git7
http://www.iss.net/security_center/reference/vuln/IPv6_Linux_Jumbogram_DoS.htm
Fragmented packets are inherently unreliable and DoS prone as resources have to be reserved until the full sequence is available to present to the consuming applications.
Note in IPv4 you could read individual fragments from the wire, IPv6 enforces the IP stack to only present fully reconstructed payloads.