On Thu, Jul 27, 2000 at 02:44:54PM -0700, Wes Hardaker wrote:
> Here's the very beginnings of a krb5 packet decoder. There are boat
> loads of problems and bad bad habits in it, but it does work.
>
> Since I was copying mostly by example without knowing what I was
> doing, do use at your own risk ;-).
>
> I'll probably clean it up more in the future at some unknown time. If
> anyone else wants to play with (ie, fix) it, however, I thought I'd
> post it now at least.
So does that mean "it's probably in good enough shape to check in with
at most some minor tweaking" (as long as it correctly decides enough of
Kerberos to be interesting, and isn't known to dump core or spew out
lots of debugging messages on packets it can't handle, it's arguably
good enough to check in), or "you probably want to do a fair bit of work
on it, or wait for the future cleanup, before checking it in"? (I have
no Kerberos traces with which to test it.)
> #ifdef linux
> #include <dlfcn.h>
> #endif
That's not necessary - that's in the SNMP dissector because it does some
run-time linking hackery to work around the binary compatibility
problems in UCD SNMP 4.1.1, but, as the Kerberos dissector doesn't call
"dlopen()" and company, it's not needed there.
> #include "etypes.h"
> #include "packet-ipx.h"
Those probably also aren't needed - Kerberos probably doesn't need to
know Ethernet type values or IPX socket values, unlike SNMP, which can
run atop raw Ethernet or IPX and thus needs to register its Ethernet
type and IPX socket values with the Ethertype and IPX dissectors.
