The MPLS WG Archive

["S.Matsushima" <satoru@ft.solteria.net>]

At 08:46 04/03/08 +0000, Adrian Farrel wrote:
> > At 14:22 04/03/07 +0000, Adrian Farrel wrote:
> > >5. 3.10.1 has
> > >       (2) At an intermediate LSR, accounting of traffic through
> > >           LSPs for each pair of ingress to egress.
> > >     This is interesting. If the label stack has just one label then I
> > >     guess we can peep inside (if the payload is IP). Is this what
> > >     you intended,
> >
> > No.
>
>I am glad to hear it.
>
> > >  or is this requirement really only limited to
> > >     "non-merging" LSPs such as LSP tunnels?
> >
> > Not limited.
> > SP uses merging LSP and need per-lsp basis accounting, they will do it
> > over merging LSP.
>
>But downstream of a merge point you cannot tell which ingress the data 
>came from.

Yes.
Therefore, it should be a requirement.
More clearly, if the LSP operated as merged end-to-end, which ingress the 
data came from
is only needed in the point of per-lsp accounting router(maybe intermediate-lsr).



>Re-reading this, I suspect you mean "incoming segment" or "upstream 
>segment" where you
>have written "ingress".
>
>Adrian

Of course I don't mean "segment".



--
Satoru Matsushima