NSP ... New Information

Wed Jun 11 05:41:06 UTC 1997

Eric Kozowski writes...

> It has nothing to do with available IP address space.  It is purely the
> number of routes in the "global" BGP routing table.  Doing rout filtering
> a la Sprint is one way to enforce better aggregation and cut down on route
> table size.

But it has everything to do with your network size.  And those who have
worked hard to keep their network size small get screwed.

Should it be network size based?  I say it should not be.

Next question:  will Sprint and other providers unfilter routes smaller
than their policy *IF* you connect to THEIR network but use that smaller
net you got from another provider?  I want to figure out just what the
actual impact of being a poorly routed member of an aggregate really
will be.

> Also, in general, the longer the prefix, the more a route tends to flap.
> Route filtering is one way to help cust down on the number of BGP updates.

Why is that?  Is there a technical reason, or is it just that longer prefixes
tend to be associated with unstable networks?

> >Are any other ISPs doing this filtering?
> 
> Yes.

My point remains, why is it that the criteria on deciding which routes
are worthy of propogating over the net, and which ones are not, has to
be based on the network size ... or how much they are wasting?  This
kind of thing certainly will end up invoking some federal investigations
here in the US.  This kind of practice doesn't let newcomers into the
business of providing nationwide or worldwide services, since such a
newcomer can't get working multi-homing without using/wasting a huge
chunk of IP space, and can't get lots of business to make legit usage
of such a space without the multi-homing.  Whether the big boys intended
to cut out the little guys or not, the impact is the same, and the
investigations will be the same (maybe the conclusions might be different,
but we obviously have people crying foul and in many cases citing federal
statutes).

Solutions?  I don't have them.  I haven't been around this stuff long
enough to know all the ins and outs.  It _seems_ to me that 45,000 routes
should be manageable with good code and proper resources.  But clearly
unabated that number won't hold at all.  The only real solution I see is
find/develop a better routing algorithm and abandon BGP4.

-- 
Phil Howard KA9WGN   +-------------------------------------------------------+
Linux Consultant     |  Linux installation, configuration, administration,   |
Milepost Services    |  monitoring, maintenance, and diagnostic services.    |
phil at milepost.com +-------------------------------------------------------+