Re: [mpls] MPLS_RT review of draft-mirsky-mpls-residence-time-06.txt

["Osborne, Eric" <[e-mail]>]

...
trimmed, one comment
.... 
> Section 4 talks about LDP and ECMP and the problems it causes.  It suggests
> that one solution is to disable ECMP used for RTM LSPs.  This seems like it
> would only make the problem worse; the more you separate the OAM path
> from the actual data path, the less useful your measurements are.  You may
> be far better off just saying "don't do RTM with MP2P LSPs".
> GIM>> You're absolutely correct, value of RTM, like of all OAM, is when "in-
> line" rule been ensured. PTP is bi-directional and must be co-routed. If IP
> paths between A and B, when ECMP disabled, considered co-routed, then
> wouldn't LDP LSPs between A and B be as well co-routed?

I can't imagine disabling EMCP and running LDP.  That takes so much out of the network operator's toolbox.  But your customers may have different needs.
Even if you disable ECMP in LDP/IGP, you still have LAGs and other technlogies.  Are you going to outlaw those too?  Again, what you want to do may work for the people you want to do it to, but I've never worked in or with a network that could get away with single-path routing.




eric