RE: WG LC: draft-ietf-l3vpn-2547bis-mcast-bgp

["NAPIERALA, MARIA H, ATTLABS" <[e-mail]>]

Yakov,

Regarding

> The text from 14.1 you quoted above clearly states that a PE
> learns that the source is no longer active "by using the same
> mechanism by which the PE learned that the source was active".
> 

I understand it. What you mean I assume is that if C-RP-PE, for example,
learned about the source via receiving Register message, it will know
that the source is no more active by not receiving Register messages.
However, the point I was trying to make when comparing C-RP-PE to plain
ASM RP was a little different and it might not have been clear:
Assume that some sources (C-S, C-G) are constantly active but they have
no receivers in MVPN (these could be some illegitimate or misconfigured
sources/end-user PCs that can be active for long periods of time, and
there could many of them). In plain ASM multicast in such cases, the RP
would keep sending Stop-Registers, the first-hop router would keep
sending Registers, and nothing else happens. In MVPN case according to
section 14, the C-RP-PE would also keep sending Stop-Registers. However,
in addition to this, the C-RP-PE will initiate for each such source a
Source Active A-D route in the MVPN. These Source Active A-D routes
might not get withdrawn for very long periods of time. So, it would be
very desirable for the C-RP-PE to announce only those C-flows that have
interested receivers.

Maria