Closed RPRX closed 1 month ago
xmux改成指针类型然后 if Xmux==nil不就行了)
xmux改成指针类型然后 if Xmux==nil不就行了)
据说有些面板会填 0
@Fangliding 等正式 release 后,文档改为“忽略或全填 0 时会取这些默认值”
@Fangliding 等正式 release 后,文档改为“忽略或全填 0 时会取这些默认值”
都已经要求面板得用ssh tunnel进行http连接了这份上了 还迁就这些行为是否有些
都已经要求面板得用ssh tunnel进行http连接了这份上了 还迁就这些行为是否有些
可能有些客户端也会这么干,总之想要 infinity-reuse 的话 maxConnections 填 1 就行了,不用留着“全 0 即 infinity”的语义
感觉现在平均 96 个子连接就开一条新连接,频率有点高,要不要改成 192 或 384
doesn't it create a pattern/fingerprint for splithttp?
The actual behavior is random within a considerable range, and we still recommend users to fill in according to their own needs, but this default value should not cause additional fingerprints
doesn't it create a pattern/fingerprint for splithttp?
这是个好问题,我还没有解释,我特地选了两个在 TLS 外不太能看出来的选项即 maxConcurrency 加 cMaxReuseTimes(而不是 maxConnections 加 cMaxLifetimeMs),并且这两个选项的默认值都是 range 取随机,最大程度上消除了潜在特征。
Lower traffic = less connections count to the same server Higher traffic = more connections count to the same server Connection count will change by the traffic rate Am i wrong? how TLS gonna hide this pattern
It is very normal for high traffic to have more connections.....
Lower traffic = less connections count to the same server Higher traffic = more connections count to the same server Connection count will change by the traffic rate Am i wrong? how TLS gonna hide this pattern
我选择 maxConcurrency 而不是 maxConnections,就是为了防止连接数成为 fixed pattern,当然如果你喜欢后者也可以手动设置
这样可以改进默认配置下 SplitHTTP 的使用体验,看了下代码,加到配置解析这里最合适,不过 API 可能会绕过这些默认值
看了下代码,splithttpNewRandRangeConfig 的 nil 改为 0 不会有问题