The MPLS WG Archive

[Bora Akyol <akyol@pluris.com>]

I think one can only overload those 3 bits so much. As we keep on
overloading these bits, we create more and more state in every LSR on
the way which is something not desirable.

I do agree that the ECN section should be removed completely from this
document.

Bora



Eric Gray wrote:

> George,
>
> One last call comment on MPLS Diff Serv (editorial
> comments via separate posting).
>
> Explicit Congestion Notification
> ================================
>
>         This section cannot remain as is if the document is
> to be allowed to complete last call.
>
>         The reference [MPLS-ECN] is no longer available (the
> '00' version was replaced by a '01' version which has since
> expired).  Perhaps a better reference would be:
>
> "ECN Interactions with IP Tunnels", S. Floyd, K. Ramakrishnan
> (http://www.ietf.org/internet-drafts/draft-floyd-ecn-tunnels-00.txt)
>
> Because there is no referent for MPLS-ECN, the text in section
> 11 ("Explicit Congestion Notification") is not workable - since,
> in the absence of work in progress on MPLS-ECN, the following
> things remain unclear (with no obvious relief in sight):
>
> o       How many bits are needed for ECN ([ECN] uses 2 bits,
>         the current text claims that a single bit is used -
>         but there is no referent from which to validate this);
> o       What the position of any ECN bit(s) would be within the
>         EXP field;
> o       To what packets ECN bits would be applied;
> o       What the implications with respect to DiffServ treatment
>         are for ECN marked MPLS encapsulated packets.
>
> This is by no means intended to be an exhaustive listing of the
> remaining issues with interactions between DiffServ and MPLS-ECN.
>
> Perhaps what is really needed is the ability to configure both
> a set of EXP to PHB mappings and a EXP field mask to be used
> to extract the EXP bits that apply to this mapping.  For example,
> if the bit arrangement is like this:
>
>       E P P
>       C H H
>       T B B
>         1 2
>
> . . . X Y Y . . .
>
> then the mask would be '0 1 1' and values 'Y Y' would be used to
> determine the EXP to PHB mapping that applies to this packet.
>
> In addition, it might be a good idea to configure - or define
> signaling for - the behavior that applies to the masked out
> bit(s).  Presumably, the default would be "forward unchanged".
>
> This section - as is - should be discarded (or possibly replaced
> with a more generalized approach - such as suggested above).
>
> --
> Eric Gray
>
> George Swallow wrote:
> >
> > This message marks the beginning of a last call on
> > draft-ietf-mpls-diff-ext-05.txt, MPLS Support of Differentiated
> > Services.
> >
> > The last call ends June 26, 2000 at mid-night GMT.
> >
> > ...