The MPLS WG Archive

["wu min" <made_in__china@hotmail.com>]

Hi,

>Roger Williams wrote:
>Tze, you may be correct, but Cisco says otherwise, at least in their
>training material. The LFIB is derived from the LIB, it is a subset of
>the LIB. Also the LFIB contains no IP information whereas the LIB does
>hold IP information. Cisco says the LFIB is generated when the actual
>shortest path to a destination is chosen. Hence, in the LFIB there
>will be one label pair for a destination (or FEC), whereas the LIB
>holds all possible paths to the destination.
>
>All this is based on the MPLS Tech Essentials, AMVS, MPLS
>Implementation, and VPNSC courses as issued by Cisco. Do I have any
>*real* proof? Not really, I'm just passing on what they say and what
>shows up in the various "show.." commands.


Yeah, I must admit that you are right about it. From the way RFC3036 
explains about LIB in section 2.7., it seems to agree with what you have 
just mentioned as it pointed out that when an LSR operates in downstream 
unsolicited mode, it would keep a collection of label binding info that it 
has learned in the LIB. When a label is needed, it will first consult the 
LIB to retrieve one to be used for forwarding (which I presume that it meant 
LFIB). I believe LIB is only significant when operating in such mode with 
liberal label retention mode combined.

I also agree with Johannes that no clear defination of such terms are widely 
available. It is also not listed under MPLS Acronym Dictionary in 
http://mplsrc.com/dictionary.shtml web page.

Thanks for this constructive argument.

-Tze Ven.

_________________________________________________________________
Chat with friends online, try MSN Messenger: http://messenger.msn.com