Hi there from a newcomer,
In the process of building my own Python AMF/RTMP client, I have decoded the content (or lack thereof) of the first handshaking packets sent in a RTMP session. These packets are mentioned at http://www.osflash.org/rtmp/protocol#handshake and http://www.osflash.org/rtmp/protocol#fn1 As has already been discovered, the contents of these packets do not seem to make any difference in the operation of the RTMP session - indeed, the version of the Red5 server I downloaded (0.6 RC1) does not attempt to match these packets to the ones generated by Flash Media Server, instead sending null bytes. However I thought this may put some curious minds at ease, or perhaps bring the protocol slightly closer to that of FMS's, even if it isn't beneficial in any great regard :) Hopefully this is the correct place to post such a thing. The contents (excluding the initial 0x03 channel ID byte) are: 0x000 - 0x003: current system uptime in milliseconds as a network-ordered unsigned int 0x004 - 0x007: seemingly always 0 bytes 0x008 - 0x600, the even bytes: the values f(1) to f(764) given by the following recurrence relation: f(0) = system uptime in ms as in 0x000 - 0x003 f(n) = (f(n - 1) * 0xB8CD75 + 1) & 0xFF 0x008 - 0x600, the odd bytes: from what I can tell by observing the Flash Player in a debugger (the way I decoded this pattern), these bytes are never touched from allocation to sending over the network, and are hence effectively undefined. I just set them to zero. Here is a simple Python routine to generate the handshake, given an uptime in ms: ----- import struct def gen_handshake(uptime_ms): handshake = struct.pack("!I", uptime_ms) + struct.pack("!I", 0) x = uptime_ms for i in range(0, (0x600 - 8) / 2): x = (x * 0xB8CD75 + 1) & 0xFF handshake += struct.pack("!H", x << 8) return handshake ----- One can write a similar routine to check handshakes from clients (ignoring the appropriate odd undefined bytes mentioned above) if so desired. I hope this is of some interest or use to someone, even if it does nothing for the project as a whole :) -- Matt D _______________________________________________ Red5 mailing list Red5@osflash.org http://osflash.org/mailman/listinfo/red5_osflash.org
