Re: [mpls] Last Call: <draft-ietf-mpls-in-udp-04.txt> (Encapsulating MPLS in UDP) to Proposed Standard

[Mark Tinka <[e-mail]>]

On Monday, January 13, 2014 10:04:13 PM Curtis Villamizar 
wrote:

> Perhaps someone needs to explain to you what a FEC is and
> how MPLS traffic is routed in most LDP networks.

Thanks for the 101, Curtis, but I think I'm doing alright 
:-).

Again, my input is rooted in operational situations that 
I've experienced, present and past (I'm neither a vendor nor 
a software developer):

J's implementation of LDP treats LDP much like a routing 
protocol (unlike C's implementation).

In such cases, LDP-derived routing has a discrete routing 
table different from the global routing table. Under certain 
conditions (design or code), there could be a mis-alignment 
between what LDP thinks is the best path vs. what the IGP 
thinks is the best path.

A knob exists within J's LDP implementation to synchronize 
LDP and the IGP (i.e., the LDP and IGP metric is always the 
same). Such a knob does not exist (AFAIK) with C because C 
don't treat LDP like a routing protocol. Without this LDP 
knob enabled in J's implementation (which is the default 
case), LDP routes always retain a metric of 1, regardless of 
the route and regardless of what the IGP metric is.

I concede that this case may not be the norm when taking all 
vendor implementations into account, but then again, my 
experience is restricted to C and J.

Mark.