@Xymmh 国内大厂基本都有境外的 CDN ，还好吧

我用的 unbound 自建递归 dns

有公网 v4 的运营商一般都会给改的，毕竟这是卖点之一
我这宽带装完，会主动问要不要改成桥接

@xy323527470 普通线路+分流也行，算是廉价 BGP 。出站靠 ip 分流，入站靠 dns 分流。

@hunono 拉 BGP 专线，给三大运营商交够钱，就没有跨网问题了

我不喜欢 22 端口，第一件事就是换掉它

随便找个 vps 和家宽建个 GRE 隧道，能通就没问题
GRE 是不能过 NAT 的

我这边配了前缀 hint 之后，ipv6 几乎变成静态的了，几个月都没换过

我有两个宽带，国内互联电信确实比移动延迟要低一点点
但移动的国际方向，尤其是亚太非常好，特别点名 v6 ，远远远远优于电信
至于美国方向，我就不清楚了，延迟太高用不到

------------------------ipv4------------------------
电信国内：
PING 223.5.5.5 (223.5.5.5) from vwan_ct: 56(84) bytes of data.
32 bytes from 223.5.5.5: icmp_req=1 ttl=50 time=4.27 ms
32 bytes from 223.5.5.5: icmp_req=2 ttl=50 time=4.61 ms
32 bytes from 223.5.5.5: icmp_req=3 ttl=50 time=4.55 ms
32 bytes from 223.5.5.5: icmp_req=4 ttl=50 time=4.29 ms
--- 223.5.5.5 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3s

移动国内：
PING 223.5.5.5 (223.5.5.5) from vwan_cm: 56(84) bytes of data.
32 bytes from 223.5.5.5: icmp_req=1 ttl=51 time=6.88 ms
32 bytes from 223.5.5.5: icmp_req=2 ttl=51 time=6.89 ms
32 bytes from 223.5.5.5: icmp_req=3 ttl=51 time=7.00 ms
32 bytes from 223.5.5.5: icmp_req=4 ttl=51 time=6.82 ms
--- 223.5.5.5 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3s

电信海外：
PING 8.8.8.8 (8.8.8.8) from vwan_ct: 56(84) bytes of data.
32 bytes from 8.8.8.8: icmp_req=1 ttl=107 time=177.17 ms
Timeout
32 bytes from 8.8.8.8: icmp_req=3 ttl=107 time=176.90 ms
Timeout
--- 8.8.8.8 ping statistics ---
4 packets transmitted, 2 received, 50% packet loss, time 9s

移动海外：
PING 8.8.8.8 (8.8.8.8) from vwan_cm: 56(84) bytes of data.
32 bytes from 8.8.8.8: icmp_req=1 ttl=112 time=37.29 ms
32 bytes from 8.8.8.8: icmp_req=2 ttl=112 time=37.28 ms
32 bytes from 8.8.8.8: icmp_req=3 ttl=112 time=37.16 ms
32 bytes from 8.8.8.8: icmp_req=4 ttl=112 time=37.18 ms
--- 8.8.8.8 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3s

------------------------ipv6------------------------
电信国内 v6:
PING 2400:3200::1(2400:3200::1) from vwan_ct: 56 data bytes
64 bytes from 2400:3200::1: icmp_seq=1 ttl=49 time=12.7 ms
64 bytes from 2400:3200::1: icmp_seq=2 ttl=49 time=12.8 ms
64 bytes from 2400:3200::1: icmp_seq=3 ttl=49 time=12.7 ms
64 bytes from 2400:3200::1: icmp_seq=4 ttl=49 time=12.7 ms
--- 2400:3200::1 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3006ms
rtt min/avg/max/mdev = 12.655/12.726/12.789/0.049 ms


移动国内 v6:
PING 2400:3200::1(2400:3200::1) from vwan_cm: 56 data bytes
64 bytes from 2400:3200::1: icmp_seq=1 ttl=50 time=16.0 ms
64 bytes from 2400:3200::1: icmp_seq=2 ttl=50 time=16.0 ms
64 bytes from 2400:3200::1: icmp_seq=3 ttl=50 time=15.8 ms
64 bytes from 2400:3200::1: icmp_seq=4 ttl=50 time=16.0 ms
--- 2400:3200::1 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3003ms
rtt min/avg/max/mdev = 15.772/15.922/15.997/0.088 ms

电信国际 v6:
PING dns.google(dns.google (2001:4860:4860::8844)) from vwan_ct: 56 data bytes
64 bytes from dns.google (2001:4860:4860::8844): icmp_seq=1 ttl=110 time=246 ms
64 bytes from dns.google (2001:4860:4860::8844): icmp_seq=2 ttl=110 time=246 ms
64 bytes from dns.google (2001:4860:4860::8844): icmp_seq=3 ttl=110 time=246 ms
64 bytes from dns.google (2001:4860:4860::8844): icmp_seq=4 ttl=110 time=246 ms
--- dns.google ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3003ms
rtt min/avg/max/mdev = 245.773/245.912/246.066/0.116 ms

移动国际 v6:
PING dns.google(dns.google (2001:4860:4860::8888)) from vwan_cm: 56 data bytes
64 bytes from dns.google (2001:4860:4860::8888): icmp_seq=1 ttl=113 time=35.2 ms
64 bytes from dns.google (2001:4860:4860::8888): icmp_seq=2 ttl=113 time=35.3 ms
64 bytes from dns.google (2001:4860:4860::8888): icmp_seq=3 ttl=113 time=35.3 ms
64 bytes from dns.google (2001:4860:4860::8888): icmp_seq=4 ttl=113 time=35.3 ms
--- dns.google ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3004ms
rtt min/avg/max/mdev = 35.199/35.265/35.299/0.039 ms

电信 v4 好，移动 v6 好
电信美国，移动亚太

移动的宽带和移动用户量，比电信联通两家加起来还要多
只能说“农村包围城市”这套理论还是更适合国情的

移动和电信互访差，不是双方的问题吗

ospf+mpls+bgp 都没有，不复杂

@kenX

建 api 中转是最方便好用的

一条千兆不如两个 500

普通的运营商，漫游用户比例很低的
一个漫游用户+几十个本地用户用同一个 IP

因为爱快的 ipv6 支持太烂了，前段时间刚换成 ROS 。

@wy315700 路由器和 ap 分开比较好

ikuai 的 ipv6 实现是个半成品，还收费