Re: [mpls] [Bier] WGLC: draft-ietf-bier-mpls-encapsulation-06

My initial reaction was same as Eric's. You either have outer stack =
tunnel in which case what's inside is irrelevant (well, we know about deep
inspection hacks but that doesn't apply here) or the label/tag is BIER. if
the label is BIER it's processed by definition by a BFR that uses BIER
entropy ...

I have no opinion whether e.g. BIER entropy should be copied onto outer
stack in case BIER is tunneled in MPLS and I consider it an implementation
choice really since I don't see a need for the architecture to mandate this
...

--- tony

On Mon, May 8, 2017 at 8:29 AM, Eric C Rosen <erosen@juniper.net> wrote:

> Hi Stewart,
>
> Thanks for your review!
>
> On 5/8/2017 10:52 AM, Stewart Bryant wrote:
>
>>
>> Please could you clarify whether this document is being reviewed by the
>> MPLS WG as part of this process?
>>
>>
> I'll let the WG chairs answer that ;-)
>
> Please can you explain how the entropy system works in MPLS?
>>
>> Are you expecting the P router to look into the BIER header, or are you
>> expecting BIER capable nodes to construct an entropy structure in the outer
>> label stack.
>>
>>
> A router that is BIER-capable uses the entropy value from the BIER header
> as part of its BIER forwarding process.  This is discussed in the BIER
> architecture document.
>
> If a packet is being tunneled from one BIER-capable router to another, and
> the tunnel passes through a sequence of intermediate routers, then the
> tunnel encapsulation probably has some entropy mechanism of its own.  It
> would make sense to ensure that if two BIER packets have the same entropy
> value in their respective BIER headers, then they have the same entropy in
> the tunnel encapsulation.  However, the tunnel encapsulation is not the
> BIER encapsulation. and how the tunnel encapsulation header is formed is
> outside the scope of this document.
>
> The BIER entropy field does happen to be 20 bits long, so if the tunnel
> encapsulation is MPLS, the BIER entropy could be copied into the EL.
>
> If the latter, you have two methods available to you the EL system, and
>> the FAT PW system. The former is becoming the method of choice where
>> available, but you have the option of using the later which will work in
>> any network.
>>
>>
> This seems like a general issue about MPLS tunnels, not a BIER-specific
> issue.
>
> I find the title a little strange since the union of MPLS and non-MPLS is
>> all networks and thus you ought to cut the title after "Replication".
>>
>>
>>
> Originally the draft only discussed MPLS networks.  When a non-MPLS encaps
> was added, the title was augmented with "and non-MPLS networks".  I don't
> really see a problem with it, as it emphasizes that two encapsulation
> mechanisms are provided.
>
> Eric
>
>
> _______________________________________________
> BIER mailing list
> BIER@ietf.org
> https://www.ietf.org/mailman/listinfo/bier
>

-- 
*We’ve heard that a million monkeys at a million keyboards could produce
the complete works of Shakespeare; now, thanks to the Internet, we know
that is not true.*
—Robert Wilensky