Re: [Bier] Feedback on draft-hu-bier-bfd-02

[Greg Mirsky <[e-mail]>]

Hi Reshad,
thank you for taking the time to review and sharing your comments, much
appreciated. I'll add to answers by Fangwei, please find them in-line
tagged GIM>>.

Regards,
Greg

On Thu, Nov 8, 2018 at 11:27 AM <[e-mail]> wrote:

> Hi, Rahman
>
> Thanks for your comments.
>
> Plese see inline marked with[hfw].
>
>
> >Hi,
>
>
>
> >I just took a quick look at the draft:
>
> >1.Section 1, 2nd paragraph mentions “…detect unicast failures..”, I
> believe this  should be unidirectional?
>
> [hfw] Agree, the BFD multipoint document only defines the unidirectional
> deteck from header to one or more tails.
>
GIM>> Great catch, thank you. Will fix in the next version.

> >2.Section 1, 3rd paragraph says that active-tail is for further study.
> So when a tail detects a failure, how will the head be notified?  If the
> head does not have to be notified, what action is taken?
>
> [hfw] Ok. Actually, we are considering the active-tails solution. It  will
> be added in the next version.
>
GIM>> Our plan is to use BFD for Multipoint Network with Active Tail
specfication as the base for this mode. Since tails in p2mp BFD are in the
Demand mode the head only can query all tail and/or the particular tail
using the Poll sequence (referred as multicast and unicast Poll). Depending
on the type of the failure, the head may recieve or not the BFD control
message with Final flag set. Absence of the response in combination with
the content of the Diag field must be used to draw the conclusion on
liveleness of the path from the head to the particular tail. From here the
head may use BIER Ping to localize the defect and switch to using the
alternative path through the BIER domain by using different transient BFRs.

>
> Regards
>
> Fangwei.
> *发件人：*ReshadRahman(rrahman) <[e-mail]>
> *收件人：*[e-mail] <[e-mail]>;
> *抄送人：*[e-mail] <[e-mail]>;[e-mail] <[e-mail]>;
> *日 期 ：*2018年11月08日 11:56
> *主 题 ：**Feedback on draft-hu-bier-bfd-02*
>
> Hi,
>
>
>
> I just took a quick look at the draft:
>
>    1.
>
>    Section 1, 2nd paragraph mentions “…detect unicast failures..”, I
>    believe this  should be unidirectional?
>    2.
>
>    Section 1, 3rd paragraph says that active-tail is for further study.
>    So when a tail detects a failure, how will the head be notified?  If the
>    head does not have to be notified, what action is taken?
>
>
>
> Regards,
>
> Reshad.
>
>
>