The MPLS WG Archive

["Adrian Farrel" <adrian@olddog.co.uk>]

Thank you,

 

Your email clarifies the points well.

 

>   A--B
>       \         |--H
>        E--F--G--|
>       /         |--I
>   C--D
>   
>   Data flows from A and C to H and I. An LDP LSP is provided from B and D to G.
>   
>   1. When you say you want to account traffic
>      through an LSP for each pair of ingress
>      to egress, you mean
>      b. LSP ingress egress pairs
>         {B, G}, {C, G}
>
>   2. At which LSRs do you want to be able
>      to perform this accounting?
>      a. Ingresses B and D
>      b. Egress G
>      c. Merge point E
>      d. Transit point F
>
> Possibly all of these.
> Where it is performed is where SP wants to perform it.

 

This is what I would expect as SP requirements.

 

> B and D should perform this accounting naturally, also E (maybe it depend implementation) too.
> In this case, what this requirement is that F and G needs something to know which ingress the data
> came from.

Agreed. And it is this requirement that I was trying to expose and make clear.

 

I would appreciate it if this point could be made abundantly clear in the draft as I do not believe it is currently understood by everyone. There are obviously significant implications for the operation of a router, and possibly for the way in which packets are labelled.

 

Thanks,

Adrian