This is a toy implentation of Sweeper, which is inspired by prior outstanding works, like [6Scan]{https://github.com/hbn1987/6Scan.git}, [Yarrp]{https://github.com/cmand/yarrp.git}, [Zmap]{https://github.com/zmap/zmap.git} and so on.

Sweeper can help you fast collect IPv6 router interface address. It works like traceroute6 but in high parallelism. Its main novelity is that Sweeper can dynamically adjust its probing focuses during the process of high-speed asynchronous traceroute6.

I write it in a single golang file with relatively few lines of code. Users can customize the running script run.sh to set entire budget (how many packets do you want to use in this probing process, $10^9$ or $10^{10}$ or more?)

But not too large, because Sweeper also follows real-world operating rules. As a scanner, its capability is limited to sending 100,000 packets per second to the internet. Our endeavor is focused on enhancing the value and yield of these limited detection packets. So, it is important not to expect it to aggressively scan the entire IPv6 network's router interface addresses in a very short time.

Sweeper has tried his best!

Its capabilities include randomly exploring multiple IPv6 prefixes with different weights, i.e., the IPv6 prefixes with high historical rewards will be probed more, like reinforment learning.

The weight random sampling is implented with follows:

1. According to weights (reward rates) of all IPv6 prefixes, Sweeper build Huffman coding tree.

2. Sweeper generate randomized bit streams and follow them to random walk on the tree from root to leaves.

3. Repeat it.

Sweeper is not very complete now, please don't use him directly for production scenes. We disclose it here only to facilitate research on IPv6 network measurements