feat: add splithttp transport support #2390
Conversation
| tlsConfig.NextProtos = []string{"h2", "http/1.1"} | ||
| } | ||
|
|
||
| // Handle ECH and Reality if needed (skipped for now or requires more adaptation) |
There was a problem hiding this comment.
Actually, you can pass mihomoVMess.TLSConfig just like the transport layers in h2, which would simplify the handling of reality and ech.
| } | ||
|
|
||
| var ( | ||
| globalDialerMap map[dialerConf]*XmuxManager |
There was a problem hiding this comment.
Do we really need a package-level globalDialerMap instead of a *Client object that directly stores these values when creating *Vless and *Vmess?
There was a problem hiding this comment.
Do we really need a package-level
globalDialerMapinstead of a*Clientobject that directly stores these values when creating*Vlessand*Vmess?
obviously (ai) copied from *ray because they have no struct for a transport to hold cross-session information(
|
|
||
| c, err = vmess.StreamH2Conn(ctx, c, h2Opts) | ||
| case "splithttp", "xhttp": | ||
| c.Close() |
There was a problem hiding this comment.
Why not implement this in DialContext or DialContextWithDialer (like the gRPC transport layer)? Creating a new conn and then immediately closing it would result in confusing traffic patterns.
wwqgtxx
force-pushed
the
Alpha
branch
3 times, most recently
from
53b7462 to
edbfebe
Compare
3 participants
No description provided.