Alvaro,

On Wed, Sep 01, 2021 at 03:41:12PM +0200, Alvaro Retana wrote:
> On August 31, 2021 at 4:51:17 PM, Jeffrey Haas wrote:
> > My second concern is shorter. Section 2.3 recommends that the p2mp BFD
> > sessions use a TTL of 255 and reference the GTSM procedures in RFC 5881.
> > However, since the destination address is a multicast group and the
> > underlying PIM protocol uses a TTL of 1 for its messages, I'm not sure the
> > 255 is appropriate for this use case.
> 
> The draft originally (-05) required using a TTL of 1, but I asked Greg
> to change it because both rfc5880/rfc5881 require the use of GTSM --
> they don't talk about any exceptions, and rfc8562 doesn't mention TTL
> at all.

Pedantically, I can't disagree with you.  RFC 5881 doesn't discuss whether
the destination address of the UDP encapsulated address must be an IP
unicast address.  But I had to check, because that's not a case I've seen
before.

> I find it very hard to justify a change in the BFD requirements; the
> BFD documents specify what the BFD behavior is, independent of the
> application.  In this case, PIM just benefits from it.
> 
> I'm not saying that the requirements can't be changed -- but if we
> were going to do that we would need some more discussion in the BFD
> WG.

I had a conversation with Lenny Giuliano to try to get a sense of what other
multicast forwarding mechanisms might take advantage of GTSM.  He wasn't
aware of any.  So, we might be in somewhat new territory.

One of his points is that 224/24 is flagged for IPV4 link local multicast by
convention and long history.  This means that to a great extent my concern
is irrelevant.  A non-1 TTL should not result in propagation when sent to
the IPv4 all-pim-routers group.

For IPv6 encapsulation, it's explicitly tagged as a link-local group and is
not a consideration.

So, for purposes of this internet-draft, I think my concern is irrelevant.

That said, I think we should clarify GTSM for multicast scenarios.  It's
fine for link local but may have concerns if it's part of a group that is
propagated.

-- Jeff

Reply via email to