bad@flatlin.ka.sub.org (Christoph Badura) writes:

> >IMHO,
> >trapping errors is much more important than the (somewhat specific)
> >use of sending bad packets to test stacks.
> 
> I think that these are different things that are largely independent.
> 
> I also think that the networking code should not panic because there's
> some illegal or unexpected value in some protocol header.  This is a
> separate issue from being able to send illegal packets.

Right now, the code uses ip->ip_len to decide how many bytes to send.
I suppose if someone made the change which might allow this to do the
right thing and still send bad packets (which I just cannot understand
wanting to do with the native code and not BPF where you have total control
over the socket) it might be easier to see what you want.

--Michael