pe@diab.se (Per Elmdahl) (10/30/89)
I have studied Inside AppleTalk and the KIP code, and I have a few
questions about the KIP protocol:
1. When a DDP packet is converted to an UDP packet, its socket
number is converted to an UDP number. Thus, a total of 254 UDP
numbers are occupied for AppleTalk traffic. Why not encapsulate
ALL DDP packets in UDP packets with ONE UDP number, decapsulate
the packet at the receiver (UNIX host) and then look at the
DDP protocol type and socket number for higher layers? This would in
my opinion be much cleaner. ("Efficiency" reasons?)
2. The UDP numbers for the dynamic sockets are about 16k. These
numbers are dynamically allocated. When starting the AppleTalk server
on the UNIX machine, you can never be sure the port is free. Why is
this scheme used?
3. There is a aaXXXXX protocol (Ether/IP/UDP/Port 901), which is
used for configuring, starting, and debugging the KFPS. There is
also aaXXXXX packets for routing and zone information, thus doubling
the RTMP and ZIP protocols. Why is this done? Why not using AppleTalk
RTMP and ZIP directly?
4. The UDP port 902 is a "rebroadcast" port. What is this port used for?
RTMP directed broadcasts?
The KIP protocol seems VERY ugly to me. Is it ugly, or do I not understand
the code? Is there a description of the protocol somewhere, or is "the code
the documetation"?
--
Per Elmdahl pe@diab.se SM5OCI DDDD /\
DIAB DATA AB Phone: +46 13 110940 D D I AA BBB / \
Box 676 Fax: +46 13 133446 D D I AAAA B--B / / \
S-582 24 LINKOPING, SWEDEN DDDD I A A BBB /__/__/\