3 ms·
With all the good and smart people over there, and considering that they have like <100 servers self-hosted, I really hoped they would be ready for IPv6 and se
by ramshanker 9y ago
With all the good and smart people over there, and considering that they have like <100 servers self-hosted, I really hoped they would be ready for IPv6 and set standards for others.
Nick: If you are reading this, any idea, when you guys would be ready. Wold be interesting to read the execution blog post.
http://ready.chair6.net/?url=stackoverflow.com http://ready.chair6.net/?url=stackoverflow.com
- skrebbel 9y agoI'm a networking noob. Why would they invest time in this? What's the benefit for them to run on IPv6? What's the benefit for you, a user?
- y4mi 9y agoipv6 clients cant talk with ipv4 servers. so they need a packet translation, which increases the latency. there shouldn't be any benefits in this particular case though, because endusers would never actually talk with the application servers. the loadbalancers are probably the only public facing servers, which probably (didn't check) support both. so communication is like this _client_ <== ipv6 ==> _loadbalancer_ <== ipv4 ==> _application server_ vs. _client_ <== ipv6 ==> _loadbalancer_ <== ipv6 ==> _application server_ the situation the GP was talking about is this: _client_ <== ipv6 ==> _ipv6 to ipv4 translation proxy_ <== ipv4 ==> _loadbalancer_ <== ipv4 ==> _application server_ you could theoretically reduce minor latency by just addressing application servers directly _client_ <== ipv6 ==> _application server_ but thats a terrible idea from both reliability and security perspectives