Yet Another Golang TLS Tunnel / 又一条 Golang TLS 隧道

[anytls]

https://github.com/anytls/anytls-go

Documentation is only available in Chinese now.

EN

In these years, the popular Golang TCP proxy in China was detected because of the "TLS in TLS" problem. Now, I want to revive this technical route.

I don't know if this is the right place to advertise, but this project is completely open source and at least I think there are some innovations.

Imitative behavior patterns always cause problems. In the current context, behavior patterns mainly refers to the padding strategy (like XTLS-Vision has a hard-coded padding strategy). Therefore, I designed a mechanism that can easily change the padding strategy.

If I were more radical, I should use a scripting language to dynamically determine the strategies for splitting and filling, but this would require the introduction of a huge script interpreter and would cause performance problems.

Because TLS adds latency to the handshake, I also implemented connection reuse.

CN

这些年国内流行的 Golang TCP 代理因为 “TLS in TLS” 问题被检测，现在我想复活这条技术路线。

我不知道这里是不是合适的宣传场所，但这个项目是完全开源的，至少我认为有一些创新。

模仿的行为模式总是会带来问题。在当前语境中，行为模式主要指填充策略（如 XTLS-Vision 就有硬编码的填充策略）。因此我设计了一种可以轻松更改填充策略的机制。

如果我更激进的话，我应该使用脚本语言来动态确定分割和填充的策略，但这需要引入庞大的脚本解释器，并且会导致性能问题。

由于 TLS 增加了握手的延迟，因此我还实现了连接重用。

[wkrp]

It is good to see development on the subject of programmable traffic shaping.

A while back I posted some strawman "challenges" for tunnel traffic shaping. For a long time I have been meaning to elaborate this idea. Does anytls have the flexibility to implement strategies like the ones posted here? (Either inside or outside the tunnel is fine.) What would it look like?
#281 (comment)

There's some other discussion about padding here: #9 (comment)

[anytls]

@wkrp I'm not planning on adding any "sleep" or "wait for server response" functionality at this point, as that would make the handshake latency very high.

Also due to performance issues, I did not implement any "all in one" multiplex like NaiveProxy. My expectation for the real-world performance of AnyTLS is that it should perform like Trojan, or better.

[RPRX]

有更多的人正视并尝试解决这个问题总归是好事，其实 Vision 也有 Seed XTLS/Xray-core#3260 ，只是 Xray 现在主要在解决其它问题

It is always good to have more people facing and trying to solve this problem. In fact, Vision also has Seed XTLS/Xray-core#3260, but Xray is now mainly solving other problems.