bpf, ipfw and before-and-after

[Barney Wolff]

On Wed, Aug 06, 2003 at 09:54:11AM +1000, Edwin Groothuis wrote:
> On Tue, Aug 05, 2003 at 10:31:01AM -0400, Barney Wolff wrote:
> > Seems to me that with ipfw logging and tcpdump packet selection this
> > is largely a non-issue.  We should be wary of adding complexity to
> > what's already at the limits of human comprehension.
> 
> Could you explain that first line a little bit more verbose?

ipfw can log packets, giving source ip/port, dest ip/port, proto
and interface which is at least some of the info that tcpdump would supply.

tcpdump can take quite complex selection criteria to determine whether
to log a packet.  So your complaint that tcpdump logs stuff that ipfw
is going to drop can be substantially mitigated.

> About the second one, given the fact that I could find out how it
> works (more or less) and where to add the statements, makes me think
> that despite the complexity of the thing being achieved, the
> implementation in the code is pretty neat and structured.

The issue is not that the addition would be complex, but that every
addition to a system already very complex must be carefully weighed
against the claimed benefits.  Does the expression "creeping featurism"
sound familiar?  Every feature of the Win32 API, OS/360 and the US
Federal Tax Code was added because somebody thought it was a good idea.

"Perfection in design is achieved not when there is nothing left to add,
but when there is nothing left to take away."

-- 
Barney Wolff         http://www.databus.com/bwresume.pdf
I'm available by contract or FT, in the NYC metro area or via the 'Net.