Re: raw socket, IP_HDRINCL byteorder

>Yah, both input and output should deal with fully assembled packets.

I'm not sure I buy requiring full packets on output.  This restriction
would have meant that e.g.
http://www.martnet.com/~johnny/exploits/microsoft/win95ping.c
would have to use BPF when raw sockets would do.

Pros:
 - Symmetry - packets get reassembled on input, so full packets on output
   make sense.

Cons:
 - Prevents certain useful uses of raw sockets (e.g. testing other systems'
   reassembly behavior).

  Bill

---------------------------------------------------------------------
The BSD APIs Discussion Mailing List
To unsubscribe:
send "unsubscribe bsd-api-discuss" to majordomo-DZEk9q7Sfju/3pe1ocb+swC/[email protected]