The MPLS WG Archive

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

Hummel,

What is wrong with the longest prefix match? besides how do you setup the labels for the control plane?

Yours,
-Shahram

> -----Original Message-----
> From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
> Sent: Friday, December 01, 2000 8:40 AM
> To: Shahram Davari; Hummel Heinrich; 'George Swallow'; 
> Dawkins, Spencer
> Cc: mpls@UU.NET
> Subject: AW: New MPLS charter 
> 
> 
> Shahram,
> A label that pilots the packet (which is e.g. a Label_Request 
> message) to the next hop router rather than by longest prefix match.
> 
> To make it clear: 
> 
> Such kind of label hasn't been specified by any MPLS protocol yet ! 
> You are absolutely right: So far, Control-plane messages are 
> not sent as labeled packets, rather they are sent as IP packets.
> 
> 
> Yours,
> Heinrich 
> > -----Urspr> üngliche Nachricht-----
> > Von:	Shahram Davari [SMTP:Shahram_Davari@pmc-sierra.com]
> > Gesendet am:	Freitag, 1. Dezember 2000 14:15
> > An:	'Hummel Heinrich'; 'George Swallow'; Dawkins, Spencer
> > Cc:	mpls@UU.NET
> > Betreff:	RE: New MPLS charter 
> > 
> > Hummel,
> > 
> > What is a control-plane label? Control-plane messages are 
> not sent as labeled packets, rather they are sent as IP packets.
> > 
> > Yours,
> > -Shahram
> > 
> > > -----Original Message-----
> > > From: Hummel Heinrich [mailto:Heinrich.Hummel@icn.siemens.de]
> > > Sent: Friday, December 01, 2000 4:52 AM
> > > To: 'George Swallow'; Dawkins, Spencer
> > > Cc: mpls@UU.NET
> > > Subject: AW: New MPLS charter 
> > > 
> > > 
> > > Georg,
> > >   
> > > you wrote"...we can define new ways of carrying labels where 
> > > new kinds of link layers may emerge out of the Optical work".
> > > Let me think in this direction:
> > > Let me call all so far defined LSPs "User Plane LSPs" because 
> > > they  transmit user data (in such a way that no transit-LSR will
> > > evaluate the IP Header of the user data packets).
> > > We may define as well "Control Plane LSPs": They should 
> > > transmit only messages for establishing/terminating User 
> Plane LSPs.
> > > Hereby the "Control Plane" labels should only be used to 
> > > determine the next hop( = replacement of the longest prefix 
> > > match mechanism).
> > > Each transit-LSR should evaluate the IP Header and even the entire
> > > "Control Plane" packet =MPLS-message like RSVP-PATH or a 
> > > CRLDP-LABEL_REQUEST,... 
> > > Remains the question how and when to install the NHLFEs of 
> > > the "Control Plane" labels: Well, BGP4, PIM, LDP-unsolicited
> > > downstream may be used to install them.
> > > 
> > > The benefit: The establishing of User Plane LSPs by means of 
> > > RSVP resp.  LDP-downstream-on-demand, can be done 
> > > almost as fast as transmitting user data by the User Plane LSP.
> > > 
> > > Another goal:
> > > Studying and specifying the case where a Control Plane LSP 
> > > and any of its associated User Plane LSPs only share
> > > the common endpoints but not the route between these endpoints.
> > > 
> > > 
> > > What do you think ?
> > > 
> > > Heinrich Hummel
> > > Siemens AG
> > > heinrich.hummel@icn.siemens.de
> > >  
> > > 
> > > 	
> > > 
> --------------------------------------------------------------------- 
> > > > Spencer -
> > > > 
> > > > The way I read that (and I would like to be wrong on 
> this, but doubt
> > > > that I am) is we can define new ways of carrying labels 
> where new
> > > > kinds of link layers may emerge out of the Optical work.  
> > > Also I don't
> > > > really anticipate any need in that area, since we 
> already have an
> > > > encapsulation over PPP and PPP is likely to be defined over 
> > > any thing
> > > > that emerges out of the optical work.
> > > > 
> > > > ...George
> > > > 
> > > > 
> ==================================================================
> > > > George Swallow       Cisco Systems                   
> (978) 244-8143
> > > >                      250 Apollo Drive
> > > >                      Chelmsford, Ma 01824
> > > 
>