From: Mark H. Degner (mark@xxxxxxxxxx)
Date: Mon Aug 14 2000 - 02:07:47 GMT-3
I don't want to turn this into a back and forth argument, but it's a good discu
ssion for fostering
input from other list members.
For your and Vijay's comments.. You are correct, injecting a default route out
to the Internet is
disasterous, but like any network implementation, you as the network designer a
re responsible for
making sure that doesn't happen. Most ISPs filter the routes advertised from a
customer, anyway,
because they've had plenty of experience with misconfigured customers. What I'
m talking about is a
default route injected from your ISP into your network.
And the hardware doesn't matter. The 4000s implemented at this particular cust
omer do exactly what
they need them to do. They didn't have to spend a lot of money on new 4500s or
7000s, and they work
perfectly. They wanted to advertise their network out. And the size of the ne
twork doesn't matter
either. Most companies don't need anything bigger than a /24. And since IANA
doesn't hand out
addresses in smaller than /19, that means these companies are getting a block t
hat is owned by the
ISP. Who says you can't register someone else's /24 in your AS Number, though?
Having less than full routes also lets your BGP table stabilize more quickly.
And the default
routes advertised from different links allows for redundancy when a link goes d
own. I'm not sure
what your experience is, but I've found the telco is the problem more often tha
n the ISP.
Let me address your bullet points..
1) Correct
2) Why a confederation? Why not just use a private Autonomous System number?
These can be assigned
from your ISP, and filtered out by them.
3) It doesn't matter what your network size is.. If you have a need for the fu
ll routing table,
than it doesn't matter what your public network mask is.
4) I don't follow you here.. What does being lazy or having strange management
have to do with
anything?
Your analogy for running NT on a 386.. If it does what you need it to do, why
upgrade for the sake
of upgrading?
Comments?
Mark Degner
----- Original Message -----
From: "Eugene Nesterenko" <eenest@msn.com>
To: "Mark H. Degner" <mark@degner.org>
Cc: <ccielab@groupstudy.com>
Sent: Sunday, August 13, 2000 9:14 PM
Subject: RE: BGP default network
> Sure, you're right
>
> But again - we live in the real world.
>
> In that case I'm sure that the main idea is to understand the placement and
> limitations
> of the protocols/technology/etc....
>
> In case of your scenario - the guys who use 4000 box, even in a multihoming
> environment
> are really strange. The monthly/quarterly payment for the links will be
> almost the same as the price for
> the new - better - router/memory (depends on the speed of the links and the
> model).
>
> So in that case more and more guys going defaultless.
>
> Also - using any kind of defaults in the dual-homing environment can be
> really tricky.
> That's my personal point of view.
>
> Of course I fully admit all that you noticed is true but acording to my
> personal experience:
>
> 1. people either don't need BGP at all - they just sit as the leaf node -
> mostly if they have less then /19 prefix.
> 2. people use confederations - to get some redundancy from the same ISP (the
> same problem - less then /19 prefix)
> 3. people run full (defaultless) BGP - they have /19 or more and they can
> afford to be defaultless.
> 4. people have /19 or more but they are LAZY or have STRANGE management -
> using some kind of your scenario
>
> So that's the story.
> Never seen other scenarios working out of the lab - i.e. in the real world.
> Of course - since BGP is the policy routing tool, you can make really lot of
> tricks with the prefered paths and so on...
>
>
> Of course your mileage may vary :-)
>
> Regards,
>
> Eugene
>
> PS. In priciple you can run Windows NT on a 386 computer - sure it'll work.
> But I'm not sure that you'll like that...
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 08:24:25 GMT-3