The MPLS-OPS Archive

[Spice Sylvia <falsesylvia@yahoo.co.uk>]

Hello Eric,

 --- Eric Osborne <eosborne@cisco.com> wrote: >  
>  > >I think we have already come out with solutions
> which
>  > >obviates the need for principals
>  > >like count to infinity, etc. As a consequence we
> have
>  > >now done away with a protocol like
>  > >RIP.
>  > 
>  > Can you please tell me what are the best
> solutions to
>  > such a class of problems? 
>  >
> 
> There's that Dijkstra-based stuff that's all the
> rage nowadays....
> 
> 
> 
> eric

Djikstra does something similar by using node
identifiers, but that is "incindental" as the solution
proposed by Djikstra is different from Bellman Ford.

Bellman Ford, in all it's abstractions does not
mandate  a "Tree" to be built. It behaves on the
"next-hop" equivalent principal.

Simply put, Bellman Ford says "to go there go
here(close by)" wheras Djikstra says "to go there...go
 there then there and finally here".

The "tree" is inherent in Djikstra as proposed in the
algorithm, hence we have to have a node identifier to
build the tree to identify points in the tree.

Bellman Ford seems like a better solution as the
complexity of the tree is not involved, and node
identifiers may not be needed.

The problem one comes across in such scenarios when
using Bellman Ford however is route propogation, which
I am sure you are aware of.

Hence the question came back to RIP, which is the 1st
Bellman Ford protocol that pops into my head. :)

I wanted to know if there are other solutions to Count
to Infinity other than giving "node identifiers" to
solve the "Bellman Ford" problems.

Hence the question on the Feedback and oscillations.

-Sylvia


	
	
		
___________________________________________________________
Yahoo! Messenger - Communicate instantly..."Ping" 
your friends today! Download Messenger Now 
http://uk.messenger.yahoo.com/download/index.html

-------
The MPLS-OPS Mailing List
Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
Archive: http://www.mplsrc.com/mpls-ops_archive.shtml