[c-nsp] filter LDP bindings

[Sergio D.]

"well, an LSR needs to allocate labels also for other nodes' loopbacks,
so this alone will not be enough ;-)"

Could it not just based is allocation of labels based on having it in the
LFIB already?

Why does the LSR need to allocate a label for all the learned prefixes?
Juniper only binds the loopback and all the LSRs only allocate labels for
that from other neighbors, maybe it just looks for /32 prefixes.
Is that what Label distribution control mode: ordered vs. Label distribution
control mode: independent is?
thanks
Message: 3
Date: Wed, 13 Aug 2008 17:29:03 +0200
From: "Oliver Boehmer (oboehmer)" <oboehmer at cisco.com>
Subject: Re: [c-nsp] filter LDP bindings
To: "Saku Ytti" <saku+cisco-nsp at ytti.fi <saku%2Bcisco-nsp at ytti.fi>>, <
cisco-nsp at puck.nether.net>
Message-ID:
       <70B7A1CCBFA5C649BD562B6D9F7ED78405DC577D at xmb-ams-333.emea.cisco.com>
Content-Type: text/plain;       charset="us-ascii"

Saku Ytti <> wrote on Wednesday, August 13, 2008 5:18 PM:

> On (2008-08-11 07:41 +0200), Oliver Boehmer (oboehmer) wrote:
>
>> BTW: the LDP filter only prevents advertisement of the binding, it
>> doesn't prevent the LSR from assigning a label (the imp-null in your
>> example).
>
> I think we had this discussion some years ago, but it would be nice,
> instead of ACLs to be able to say 'no mpls ldp label; mpls ldp label
> loop0' or so, to generate label only for loop0.

well, an LSR needs to allocate labels also for other nodes' loopbacks,
so this alone will not be enough ;-)

However, IOS now has a label allocation filter
(http://www.cisco.com/en/US/docs/ios/mpls/configuration/guide/mp_ldp_all
oc_filter.html<http://www.cisco.com/en/US/docs/ios/mpls/configuration/guide/mp_ldp_alloc_filter.html>)
having a "allocate global host-routes" shorthand to only
allocate labels for /32s (or uses a prefix-list for more granular
control)..

       oli


-- 
Sergio