[j-nsp] Redistribute Connected

[Frans Legdeur]

Dear Mark, although off-topic, just my 2 cents ...
> 
> On Tuesday 25 November 2008 13:53:38 Frans Legdeur wrote:
> 
>> Also ensure that unknown segments
>> are advertised internally, so including the /30 between
>> you and your transit neighbor, else your other router
>> doesn't know where he can find his next hop for the i-BGP
>> routes. (There are other ways to solve this, like
>> next-hop self or add the interface passive to OSPF/ISIS,
>> but this is the best way to my opinion ;-)
> 
> I think setting the NEXT_HOP attribute within iBGP, by
> default, is a best practice. I'd caution against any other
> method.
> 
> I personally don't prefer to carry foreign addresses within
> my IGP, but this topic has already been hashed out on this
> list in great detail...
> 
> Mark.

You will loose valuable information, in case you use route reflectors and
software like ARBOR, that collects traffic and routing information to
understand where all your traffic is destined to. In your case you would
only see the router as endpoint, not the real end-point that lays behind
your router.
Mostly those core routers collect a lot of BGP sessions to transit and
peering end-points. They will now become unknown to you on your internal
routers ...
This makes troubleshooting a case of hop-by-hop peeking instead of a single
point for all information.
That's what I got against it, you will hide information behind your router.

What you have against foreign  addresses carried in your IGP, is more the
way how you would do that. I'm very much against the method by adding them
as passive interface into your IGP. If you would redistribute them, they
will appear as External routes, indeed, just what they are, since they don't
belong to your IGP.

I agree, all methods work, but without arguments I can't agree that it's a
best practice ...

Kind regards,


Frans.