Re: [mpls] Question about 0x8848 Ethertype

["Shahram Davari" <davari at broadcom.com>]

Nitin,

But RFC5332 explicitly says "Ethertype 0x8847 is used whenever a unicast ethernet frame carries an
MPLS packet". 

So we can't use 0x884 for unicast frames. Can we? If so then RFC 5332 has a problem.

Regards,
Shahram

-----Original Message-----
From: Nitin Bahadur [mailto:nitinb at juniper.net] 
Sent: Thursday, October 22, 2009 12:04 PM
To: Shahram Davari; mpls at ietf.org
Subject: RE: [mpls] Question about 0x8848 Ethertype

> If we assume your interpretation of the RFC that the context label is
> unique only if it is used as upstream label, meaning that the same context
> label can be used as downsteam-assigned label on that LAN, then the
> question is what about unicast packets in a LAN? 

> How will a node receiving
> a unicast packet can identify whether the top label is upstream assigned
> or downstream assigned?

Using ether-type...0x8848 will indicate upstream assigned.


> Thanks,
> Shahram
> 
> -----Original Message-----
> From: Nitin Bahadur [mailto:nitinb at juniper.net]
> Sent: Thursday, October 22, 2009 11:28 AM
> To: Shahram Davari; mpls at ietf.org
> Subject: Re: [mpls] Question about 0x8848 Ethertype
> 
> 
> I'm not 100% sure if your interpretation is correct.
> 
> The RFC I think tries to say that the context-label must be unique. In
> other words both P1 and P2 cannot use the same label to
> denote their context . In other words, upstream distributed labels must be
> unique when using ether-type 0x8848.
> 
> I don't think the intention of the RFC was that both upstream and
> downstream allocated labels must be unique.
> 
> Nitin
> 
> On 10/22/09 11:17 AM, "Shahram Davari" <davari at broadcom.com> wrote:
> 
> Hi Nitin,
> 
> Thanks for your answer. But RFC 5332 requires the use of unique "context
> label" at the top of the label stack which can't be used as downstream-
> assigned label on the multi-access media. So the mere presence of a
> "context label" should be enough and we don't need a dedicated Ethertype
> to tell us that the context label is upstream assigned.
> 
> Do you agree?
> 
> Regards,
> Shahram
> 
> -----Original Message-----
> From: Nitin Bahadur [mailto:nitinb at juniper.net]
> Sent: Thursday, October 22, 2009 11:10 AM
> To: Shahram Davari; mpls at ietf.org
> Subject: Re: [mpls] Question about 0x8848 Ethertype
> 
> 
> See NB> below...
> 
> On 10/22/09 10:37 AM, "Shahram Davari" <davari at broadcom.com> wrote:
> 
> Hi,
> 
> RFC 5332 says:
> 
> "Ethertype 0x8848, formerly known as the "MPLS multicast codepoint", is to
> be used only when an MPLS packet whose top label is upstream- assigned is
> carried in a multicast Ethernet frame."
> 
> It also says:
> 
> "It is expected that the LSR will follow the procedures of [RFC5331],
> pushing on two labels, with the topmost label being a "context label" that
> is the same for all MPLS packets being transmitted by the LSR onto the
> Ethernet, but with the second label being different for different LSPs"
> 
> And RFC 5331 defines context label as:
> 
> "A "context label" is one that identifies a label table in which the label
> immediately below the context label should be looked up. A context label
> carried as an outermost label over a particular multi-access subnet/tunnel
> MUST be unique within the scope of that subnet/tunnel."
> 
> Considering the above statements we can conclude that:
> 
> "Multicast MPLS packets with upstream assigned labels encapsulated in
> Multicast Ethernet frames in a Multi-access media (LAN) require to use
> 0x8848 as Ethertype and require to use a context label that is unique over
> that media to identify the label space for the Multicast MPLS packet"
> 
> The question is since a context label is unique in a LAN and can be
> immediately identified as an upstream label, then why do we need 0x8848
> Ethertype? Looks like we could just use the context label.
> 
> Am I missing something?
> 
> NB> The question is "how to identify that this is an upstream label".
> Consider label 100. This label could have been allocated by upstream P1
> router as an upstream label and distributed to downstream router PE. The
> same label could have been allocated by PE router and distributed to
> router P1. So now, when PE
> receives a packet over the LAN (from P1) with label 100, it does not
> whether to look it up as an upstream label or as a downstream label..
> 
> Ethertype 0x8848 allows PE to perform lookups in context of P1 and not in
> context of labels distributed by PE.
> 
> Thanks
> Nitin
> 
> 
> 
>