38 lines
1.8 KiB
Plaintext
38 lines
1.8 KiB
Plaintext
|
|
|
|
The basic NORP server state machine is in place and pretty much ready for testing. There are some details that need to be wrapped up:
|
|
|
|
1) Implement ACK-based flow control using double-buffer markers, explict positive acknowledgement, (DONE) 1a) Add option to NORM to
|
|
only ACK after application has read stream data. (DONE)
|
|
|
|
2) Clean up debugging messages. (SOME)
|
|
|
|
3) Provide NORM congestion control options (DONE - CCE added)
|
|
|
|
4) Session configuration options (ouch!) 4a) Src/Dst based proxy methods and parameters
|
|
|
|
5) Implement auto handoff to TCP direct connnect on port unreachable (what about NORP request timeout, too?) (both DONE)
|
|
|
|
6) Use protolibd for "safer" socket state
|
|
|
|
7) norm_rx_stream can hangup shutdown - fix it (DONE)
|
|
|
|
8) Win32 build and test
|
|
|
|
8a) Win32 NORM-CCE support that's not WinPcap dependent
|
|
|
|
9) Send / handle negative SOCKS replies
|
|
|
|
|
|
10) IMPORTANT: Detect remote "norp" failure _or_ make sure that we can _always_ detect when our local TCP socket (originator client, or
|
|
correspondent remote) closes. Right now, we stop TCP socket input notification upon NORM flow control block condition (e.g. failed
|
|
watermark) ... we probably should have some failed watermark timeout condition, perhaps in conjunction with using an
|
|
application-defined command as a "keep-alive" mechanism. For example, have each NORM sender send a "keep-alive" heart beat
|
|
periodically, but upon multiple periods of failed watermark ACK _and_ missing "keep-alive", initiate session teardown? Otherwise, this
|
|
creates a condition of persistent requests of ACK indefinitely.
|
|
|
|
|
|
11) To provide easier firewall traversal, should we split the "correspondent" norm tx & rx sessions so that it can "talk back" to the
|
|
originator source port. Can also use the NORP messaging to signal the norm ports to use.
|
|
|