The MPLS WG Archive

[Shahram Davari <Shahram_Davari@pmc-sierra.com>]

Curtis,

> > > 4) What happens if in an FEC stack ping, some of the FECs 
> are confirmed
> > > and some are not? Do we need more error codes? How many more?
> > 
> > Good question.  Let me answer that one separately in 
> another message.
> > It is an ommision in the document.
> 
> There are a couple of legitimate situations where this will happen.
> With varying degrees of "legitimate".  Some have to do with TTL
> processing with hierarchy.  Another is situation is an LSR that
> doesn't decrement TTL (hopefully not on an LDP LSP that loops) or one
> that doesn't understand lsp-ping.
> 
> What you'd see at the ingress is an error.  If you resent the Echo
> Request without any of the prior mapping TLVs, then you'd see an Echo
> Reply that confirmed the FEC and gave a downstream mapping TLV.  You
> could then continue the traceroute.  Having reached the egress you
> have a portion of the trace for which you aren't certain how the
> packets go.  If the pings work and the area of uncertainty is outside
> your administrative domain, you have to accept that delivery happens
> by some undisclosed magic until delivery fails to happen.  If it is
> within your administrative domain, you can MIB query until such time
> as you can resolve the problem to get complete traces.


What you just said is true and acceptable for when some of the
labels in a Downstream mapping TLV (DM TLV) are confirmed and some are not.
In that case, sending an echo-request without DM TLV will be successful and
will give you the downstream mapping. My suggestion is changing the description
of error code 7 to:

Replying router is one of the "Downstream Routers", but its mapping for this FEC are the given label stack."

My  concern was more on pinging Traget FEC stack. Let's say one of the
FECs in the stack is confirmed but other ones are not confirmed. What is the error reporting procedure? My suggestion is changing the description of error code 4 to:


" Replying router has no mapping for one or more FECs in the Target FEC stack"


> 
> > > 5) How are the ATM/FR labels supported in Label-mapping TLV?
> <snip>
> 
> Lets try again on this one.
> 
> AFAIK, an ATM or FR LSR doesn't decrement TTL.  Therefore this would
> appear to to be a transparent couple of hops just like FSC, LSC, TDM.

Agree.


-Shahram

> 
> This might be one of those situations in point 4 above where there is
> a discontinuity in the sanity of the traceroute.  The downstream
> mapping gives the next ATM LSR but incrementing TTL does to the end of
> the VC if TTL isn't decremented.
> 
> Curtis
> 
> ps - IMHO - PHP is a thing of technical elegance and beauty compared
> to MPLS over ATM (or anything over ATM).  :-)
>