News for hackers. Stuff that matters.

A comment on IAB Statement on IPv6. Parent comment

Re:IAB Statement on IPv6 by danudey on Tuesday November 08, 2016 @03:31PM (#12903658)
Heard a "funny" story about Facebook switching over to pure IPv6 and their issues dealing with largely untested IPv6 implementations. For example, switches which, when presented with an IPv6 BGP route while they don't have IPv6 configured, crash. Apparently they took down an entire data centre full of rack switches finding that out.

My second-favorite problem, after that was solved, was developers constantly using IPv4-only code. Their eventual solution was to just disable IPv4 entirely so that anyone committing IPv4-only code was committing broken code.

It's amazing how much work it takes to bring people into the future.

[ Parent | on HN ]
Re:IAB Statement on IPv6 by eb0la on Tuesday November 08, 2016 @04:34PM (#12904143)
Very "funny", but that kind of crashes happened more than you imagine.. in telcos.

Usually the problem is memory. A machine with full routing enabled needs much more memory for IPv6 than for IPv4 and when routers run out of memory they just crash, reboot, and start again...

[ Parent | on HN ]
Re:IAB Statement on IPv6 by CountSessine on Tuesday November 08, 2016 @10:28PM (#12905852)
Usually the problem is memory. A machine with full routing enabled needs much more memory for IPv6 than for IPv4

Interesting - why? I would have thought that routing tables for ipv6 would be a fraction of the size of their ipv4 equivalents. Am I wrong? Or is this just sloppy programming on the part of those switch programmers?

[ Parent | on HN ]