All things related to multicast

[Bill Fenner <>]

>2. We really need a way to deploy multicast other than "buy Cisco
>routers."  ...
>... One possibility would be
>to have a current implementation of the multicast protocol in a computer
>that sits behind a non-multicast router

This isn't quite enough; another piece that's missing from the puzzle
is the ability to use an alternate IGP (or alternate metrics in the
existing IGP) for multicast-only paths.  Once you modify the topology
by adding multicast-only routers and/or tunnels, the multicast routing
protocol needs to learn about them.  PIM relies on the topology built
by whatever unicast protocol you're using; if that topology does not
reflect the multicast topology then PIM won't work.

It's really unclear to me why we don't have modified IGP's like we
have MBGP -- I remember hearing ISP's express this as a requirement
to Cisco over a year ago.  As it stands, Cisco only allows the use
of DVMRP or static routes to route along incongruent topologies.
DVMRP has such a bad name that nobody will use it even when it may
be appropriate, and static routes are generally insufficient (and
people don't like to configure them anyway).

>As far as I know, there is no PIM-SM/MGPD public domain software
>implementation that we could deploy

I haven't tried it yet, but I was under the impression that gated had
both PIM-SM and MBGP working.  However, yet another problem arises: many
campuses have deployed PIMv1, and gated implements PIMv2.  There's no
published document describing PIMv1; I have a stashed copy of an old spec
which might or might not reflect what Cisco does.  Lip service says that
it's not worth implementing PIMv1 since the rest of the world can just
upgrade to IOS12; reality might say something else.

  Bill