Re: [mpls] [Bier] [bier] The first nibble issue associated with MPLS encapsulation

Hi Loa,
I think that use of value 5 in the first nibble for MPLS encapsulated BIER packets is to differentiate them from all other variants that use the first nibble as the hint to P node on what use as flow characteristic information for ECMP. I don't think, though we should consult with BIER WG and authors of draft-ietf-bier-mpls-encapsulation, that the first nibble not supposed to be used as payload type identifier but only to differentiate from other types of the first nibble at P nodes to avoid incidental hashing of BIER packets assumed to be something else, IP or Ethernet. If that is the case, then the same value 5 may be used as None-of-the-Above indicator, including SFC.

	Regards,
		Greg

-----Original Message-----
From: mpls [mailto:mpls-bounces@ietf.org] On Behalf Of Loa Andersson
Sent: Thursday, April 14, 2016 7:54 PM
To: Stewart Bryant; Nagendra Kumar Nainar (naikumar); Eric C Rosen; Alexander Vainshtein
Cc: mpls@ietf.org; bier@ietf.org
Subject: Re: [mpls] [Bier] [bier] The first nibble issue associated with MPLS encapsulation

Stewart,

On 2016-04-15 03:36, Stewart Bryant wrote:
>
>
> On 14/04/2016 16:11, Loa Andersson wrote:
>>
>>
>> /Loa
>>
>> On 2016-04-14 22:41, Nagendra Kumar Nainar (naikumar) wrote:
>>> Hi,
>>>
>>> Per IANA registry, 0-1 are ³Reserved² and so is not strictly 
>>> assigned with PW. Something like a modified CW beneath LSE?. 
>>> Further, BIER is hop-by-hop lookup based. So I think we don¹t need 
>>> to include this always.
>>> Instead, we
>>> could conditionally include this, if the next BFR is tunneled (not 
>>> directly connected).
>>>
>>> The receiving node will use the label to identify the payload/FEC. 
>>> So I think, the presence of modified CW should not confuse it with PW.
>
> Not quite Loa.
>
> It could be an Ethernet PW with no CW.

Yes - you are right, how much of a problem is this in a P node? Any value, except 4 and 6, will stop the P node from doing ECMP/Load Sharing, right? A P node will never look for the control word.

So an Ethernet PW that carries the same value in the first nibble that we specify for bier will be treated the same way as a bier packet.
This does not effect the bier packets, and is the right way to treat the PW.

So where do we have the problem?

BIER would be in the clear, as it will carry any value but 4 or 6 in the first nibble, right?
PWs with control word will have 0 (or 1) and not be effected by a P node that are doing ECMP/Load Sharing.
Ethernet PWs without control word might have any value in the first nibble, if there is 4 or 6 a P node might to ECMP/Load Sharing.

It seems to me that specifying the first nibble in a bier packet would be good, for the time being I don't care if we say 0 or 5.

If there are problems with Ethernet PWs without control word, this has to be solved in the context of PWs.

/Loa
>
> Stewart

_______________________________________________
mpls mailing list
mpls@ietf.org
https://www.ietf.org/mailman/listinfo/mpls