The MPLS WG Archive

[Matt Squire <mattsquire@acm.org>]

Matt Squire wrote:
> 
> Eric Rosen wrote:
> >
> > Eric> We just need to invent a way to ensure that the UDP packet follows the
> > Eric> IP hop-by-hop  path,  rather  than  being  sent through  an  LSP.
> >
> > Yakov> or just use what is described in draft-pan-lsp-ping-00.txt.
> >
> > The point is to find a simple mechanism that does not depend on a particular
> > control plane.
> 
> To jump on an already crowded bandwagon...
> 
> Having an LSP ping that is signalling protocol dependent does suck.  But
> I agree that if they only thing you're sure about in the reverse
> direction is the reverse control path, then it might have to be used.
> 
> To this end, it would seem simple enough to compartmentalize the
> signalling aspects of the proposed LSP ping that it has some header
> stuff and then some signalling protocol specific TLV.  The RSVP TLV
> could be as described in this draft, and a CR-LDP TLV could be just as
> easily defined (maybe just the LSP ID?).  So at least the methodology
> would be consistent as well as the main packet format and port numbers,
> etc., with signalling specific junk tacked onto the end.
> 

It would also be nice if the signalling protocls behaved similarly,
maybe using some type of notification message with a PING_OK status or
something, but thats doesn't seem like as big a deal.

- Matt