smurf, the MCI-developed tracing tools (was Re: Bogus announcement)

Karl Denninger writes...

Then you damn well better not be permitting any of the following:

1) Forged source addresses (this CAN be stopped with specific filters
  on your interfaces, although some will bitch about the performance
  impact - depending on their specific choices)

Yet another case for pressing (now) ARIN (and others who allocate address
space) to do address allocation in reasonable chunk sizes instead of forcing
providers to accept little bits of address space a piece at a time. Prefix
volume on BGP would be helped, too, by having fewer little pieces scattered
all around. And with address space now on a paid basis, in theory people
will generally ask for what they expect to need (there will be exceptions
but they should be easy to spot), so there is more reason to actually give
out requested allocations that are not obviously inflated.

2) Directed broadcasts (which are used to "create" these DOS attacks by
  bouncing the attack off a particularly-well-connected location,
  USUALLY a provider's internal infrastructure).

Block both of those and Smurfs would disappear. If you can trace the TRUE
source of such an attack quickly, people will go to jail for this. The only
reason they are popular is because the source addresses CAN be forged.

Specific information is always helpful. Unfortunately, if it has been
given on NANOG, it can be missed due to the high noise level (yet another
issue we need to work on). Would config examples in IOS and gated be too
much to ask for (if someone only knows one, someone who knows the other
should follow up).


Agreed. Let's make it easy.