Re:[Bier] Flow Label [was Re: BIER in IPv6 ---draft-zhang-bier-bierin6-04]

Hi Brian,

Thank you very much for your comments!

Please find my answer below with [Sandy].

Thanks,

Sandy

原始邮件

发件人：BrianECarpenter <brian.e.carpenter@gmail.com>
收件人：Tony Przygienda <tonysietf@gmail.com>;张征00007940;
抄送人：6MAN <6man@ietf.org>;BIER WG <bier@ietf.org>;
日 期 ：2020年03月23日 12:04
主 题 ：[Bier] Flow Label [was Re: BIER in IPv6 ---draft-zhang-bier-bierin6-04]

Hi,

I have not read the whole draft but I did notice this:

>    The Flow-ID in the IPv6 packet SHOULD be copied from the entropy
>    field in the BIER encapsulation.

It's good that the BIER entropy field defined in RFC8296 is 20 bits long, but there are a few issues here:

1. Nit: in RFC 8200 the field is named "Flow Label", not "Flow-ID".
[Sandy]: Thank you for your pointing out this nit, the name of Flow Label will be corrected.

2. The text technically violates RFC8200, which states that the flow label is defined by RFC6437. 
3. RFC6437 recommends the flow label generation procedure to follow a certain rule:

   It is therefore RECOMMENDED
   that source hosts support the flow label by setting the flow label
   field for all packets of a given flow to the same value chosen from
   an approximation to a discrete uniform distribution.

but I didn't find any discussion in RFC8296 of how the BIER entropy value is generated. 

I think you could get round these two points by something like:

   The entropy value used in an IPv6 flow label SHOULD be generated
   by a method compatible with [RFC6437].
[Sandy]: Thank  you for your suggestion, the description will be updated. And you are right, BIER didn't define the generation of entropy field. There are many ways to generate the field for load-balancing or other purposes.

4. Finally, it seems to me that the flow label is useless by definition for link-local traffic; however, it's harmless. (I never really thought about that before.) Of course, it could be relevant for globally-reachable addresses.
[Sandy]: The entropy field takes effect on BIER layer. So yes, it may also be filled in flow label field, we use "SHOULD" here. :-)

Regards
   Brian Carpenter

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