« March 1997 | Main | May 1997 »
I often wonder why the largest ISP's don't using out-of-band
signalling for routing information. The dynamics which routing
protocols create are fine for small networks but can create real
havoc in a large nationwide network.
Why not calculate all the routes in one place? These days when "next
hop shortest path / reachability" style routing is no longer correct,
we need to have a central place to apply policy. The interesting
thing about policy is that is has little to no place for dynamic
rerouting. Policy is a force for rigidity.
It sure seems like the people working on tag switching are more
interested in keeping their low-capacity switches and routers alive a
little longer than in implementing policy. What the world needs is
policy because policy equals differentiated billing rates. This looks
like a train to me.
Most people don't realize it but email can be routed just like network
packets. If one publishes their email address widely it doesn't make
sense to have to change it when you move your office. That's where
email routing comes in.
I would guess that large organizations with thousands of people find
this a nightmare. It's also something not addressed by mail systems
such as Microsoft Exchange or Lotus notes. This e-mail routing is
really a layer above (or below, depending on your vantage point).
It's between the final work-group email server and the
router/firewall.
Do NAS vendors ever actually
guess is no. I'd love to know if the CEO of 3Com and US Robotics
actually use ISDN from home and call in to their office using 28.8
modems. They're probably too busy. My sense is that few people at
Shiva actually dial in from home. I'm sure the folks at Ascend don't
use their products (because if they did they'd notice all the bugs and
the fact that their boxes reboot all the time).