single or multiple range statement

[David W. Hankins]

On Mon, Jul 03, 2006 at 06:19:31PM -0700, George C. Kaplan wrote:
> Is this really a problem nowadays?

I think we all have a story of how .0 or .255 have bitten us.

I remember working at a dial access ISP, we discovered that we could
slightly downsize the subnets we allocated to each access router
for dynamic clients.  The optimization relied on using the .0 and
.255 addresses.  But it also saved a /lot/ of address space, and
we were able to put off RAM upgrades on our router core by keeping
our IGP size down this way.

We hemmed and hawed over this for a few weeks and then went ahead
and deployed the scheme (a perl script did all the heavy lifting
at midnight).

Surely, no PPP client would care, at least none that we tested seemed
to (of course, those tests didn't include the router changes, they
were done with static assignments).

That morning, on my 0500 NOC shift, we started getting phone calls
that folks could connect but couldn't get net.  About 0700, just
when our 'morning flood' of customers are waking up to check their
email while they eat their breakfast, so the dynamic ranges are
getting their first workout.

They were all on .0 addresses.  But not all .0 addresses.  Some
worked (I often though it was strange, me pinging these unsuspecting
customers), some didn't.

It turns out, if you advertise x.y.z.0/28 via RIP2, and redistribute
that into OSPF, and have a x.y.z.0/24 route already being advertised
in OSPF (to a discard interface on the IP router facing the dial
access routers, let us imagine), the RIP2 route does not get
propogated.

The OSPF implementation used the network number as a 'name' for
the route, so it couldn't support two routes starting from the
same network number.  I was never clear if this was a protocol
flaw or an implementation flaw.  I suppose for a DHCP discussion
list, it doesn't matter.

Is that a problem nowadays?  Probably not.  I doubt modern
dial access networks look anything like the sharpened sticks
and blunt stones we used back in those days.


When I was growing up, my uncle used to keep dairy cows.  They had
a simple, pulsing electric fence.  It would throw some volts down
a thin metal wire around the pasture, and the odd cow...reaching
under the wire to try and get at better grass usually...would get a
little jolt to send them back.

Some days, we forgot to turn the fence on.  But unless there were
some really young cows who hadn't been 'bitten' by the fence yet,
they never went near it.  One jolt was enough it seemed, once the
cows got their bite the fence could just as well be off.

Now my uncle, if we were out in some part of the property on some
errand, would walk right up to the fence and grab hold to test if
it was operating (if it was grounded anywhere along the line before
that point (say a fallen branch, or if the wire wore through the
insulator it hangs on), you won't get a jolt).  Fence maintenance
on a farm is a ceaseless project, and there's no better way to know
for sure.

But I could never bring myself to do that.  I think he goaded me
into doing it a couple of times just to see me wince, but I never
really wanted to.

One jolt was enough for me too.


So, would I still prefer not to use those addresses if I had a
choice?

Probably.  I've already had that bite me once, and it was enough.

-- 
David W. Hankins		"If you don't do it right the first time,
Software Engineer			you'll just have to do it again."
Internet Systems Consortium, Inc.		-- Jack T. Hankins