this doesn't make any sense. .
The stream would literally have to be flooded everywhere and only IGMP on the switches would limit the propogation. You wouldn't be able to prune because you aren't running a multicast routing protocol so each L3 device is completely unaware of any other L3 device from a multicast perspective.
But yes, anything is possible if you need to rely on theory I suppose.
Ah, I guess I'm unclear on this too. I'm unsure anything is possible in theory, but the theory I'm interested in supports real operations and/or help you understand them. Or, theory to helps one understand and make sense of how things work.
To better understand flood-and-prune, first forget about switches, think hubs connected to the router. Second, forget, initially, multiple L3 devices.
If a multicast source comes up, the connected router interface will see the multicast stream. In theory, the router could immediately flood that multicast stream to all the other router's interfaces. (I believe, that's what PIM-DM does, possibly DVMRP too.)
Additionally, router interfaces will see IGMP messages. So, it knows whether any hosts want the stream it's flooding. If there are no hosts that desire the stream, the router can prune it off the interface. (I also believe, that's what PIM-DM does; and also again, perhaps DVMRP too.)
So, on a single router, in theory, you don't need a full fledged multicast routing protocol to support multicast between connected interfaces.
Whether what I've described is truly multicast routing, is debatable, as IP networks aren't even considered, only interfaces. What I've just described is more like bridging than routing.
What I've also described could be extended between routers. Doing so creates some issues, for instance what controls pruning between routers? In theory, you could also use IGMP messages, much like some "smart" switches suppress IGMP multiple host messages to the router's interface or cascaded switch uplinks, toward the router's interface. Doing this, it would work fairly well as long as you didn't have alternate paths or multiple routers with interfaces on the same segment.
The reason I keep saying, in theory, is because just doing multicast as a special case on a singe L3 device is rather pointless if you can also do it for multiple routers, and when working with multiple routers, you'll want to address the other issues of multicast.
Why I even mention this, is because PIM-DM (and DVMRP?) is built on a basic flood-and-prune model. That, and many decades ago, when first learning multicast, I still have some very vague recollection you could get a Cisco router to do multicast between its interfaces without a full multicast configuration. Why I think I remember it at all was, because it was something I wouldn't expect to work. (Also something you shouldn't intentionally be doing.) Again, though, it was literally decades ago.
Unfortunately, I don't have the equipment to lab this up. Maybe I'll shoot Peter a note. Good chance he might know and/or be willing to lab it up.
[edit]
I just shot Peter a note, while doing so, though, I think I recall doing multicast on a singe Cisco router, without all expected multicast statements, was something I read, but don't believe I tried. Where I had read it, I trusted the source material, but I was much more trusting then. ;)
Also, this was back in the version 11.0 days, so its also possible it's something that worked on such a version, but has been "fixed".
Joe
However, if your subnets are across multiple L3 devices, you'll want some kind of multicast routing protocol also enabled
That's what PIM is ie. a multicast routing protocol.
It uses the IP routing table.
Jon
I agree with both your statements, but what's factually wrong with my statement? Why your post? What's the error?
I read your reply, as you want to make clear PIM is (also) a multicast routing protocol (which some might not know). Again, I agree, and even said so in my reply.
However, I may have muddied the waters by further saying that statement was meant to be be generic both for multicast protocols and vendors. I further compounded that muddying by addressing different kinds of PIM. (You mentioned you can be pedantic - laugh - meet your master.) If you like, I'll admit adding this information might have been a "mistake", but am I factually incorrect? (If so, please correct me.)
Your next reply had:
Joe
on the same Cisco device, I believe, PIM alone will generally be all you need. However, if your subnets are across multiple L3 devices, you'll want some kind of multicast routing protocol also enabled.
thanks for mentioning that because that specific comment of mine was for multicast, generically, not Cisco specific, while my statement just ahead of it was Cisco specific.
Sorry but that's not what you said.
There's nothing wrong with admitting a mistake, I do it all the time on these forums :-)
Jon
Now your saying that's not what I said, and I made a mistake.
Okay, but I'm still not seeing the contradiction, even though you've highlighted the words same and multiple, in two different sentences.
I think, but I'm really not sure, you think I'm saying PIM only applies to the same device? Well, as you noted, PIM is also a multicast routing protocol, and I agree (and know that too), so is there a contradiction? Or, do these statement only contradict if you don't know that?
Jon, I'm really still trying to "see" what you see is my mistake.
I readily admit, that PIM (on interfaces alone), may be insufficient for same Cisco device multicast. Again, I originally noted I believed that might be true, but I don't know it to be true (and you believe it's not).
I apologize to OP and other readers for any confusion my posts in this thread may have caused. Hopefully with Jon's assistance, I'll eventually get it right.