AW: 232-Addresses not only for current SSM model

[Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>]

	 Doron answered:

> I guess you mean 232/8 (the range 232/24 is 232.0.0.xxx).
	 Thanks. you are right. 
> I tend to agree that 16M addresses seems a lot. However, keep in mind
> that:
> 
> 1. Reverse-IGMP, or On-Demand Multicast, will cause a server to use a
> lot of addresses, about one address per URL per rate. This can be a lot.
	 Sure. But 1 million or  500 000 is still sufficient, isn't it ?

> 2. In order not to have too many Ethernet MAC layer collisions, we
> should have some bits in the address field, selected randomly by the
> server.
	 
	My point is another one or maybe I should ask this question:
	When a node in the middle of the network receives a particular "multicast" message, how can he
	recognize which  evtly. present  (S,G) -multicast channel is meant in case there are multiples of  (S,G)-models, and not just SSM ?

	Heinrich









	 
> Subject: 232-Addresses not only for current SSM model 
> 
> 
> 	Hello,
> 
> 	The 232/24-address space is reserved for source-specific
> multicast. An address G from this address space only uniquly identifies
> a multicast channel
> 	together with a particular source's unicast address S.  
> 
> 	Application: multicast model according to the SSM WG.
> 
> 	1) 
> 	This is really great for many very good reasons.
> 
> 	2)
> 	A particular source of unicast address S does not head  2**24
> SSM-multicast channels. Therefore, a subrange would do as well.
> 
> 	3)
> 	I can imagine further multicast models in the future, which
> would be happy to use a 232/24-address, but which
> 	would use different protocol procedures, protocol messages,
> protocol TLVs. 
> 	Just one example:
> 	A completely different mulitcast delivery channel could be a
> ring. As a matter of fact, a ring needs only about 20 % more hops than a
> Dijkstra-tree.
> 	The information can be sent out in two directions so that in
> case of a link failure, each receiver node would get the information at
> least once.
> 	Furthermore, in case of optical networks, the power of light is
> preserved in a better way.
> 	Of course, a ring could be extended to a "tree of rings" or a
> "ring of rings".
> 	 
> 	 I am quite sure that other folks have further ideas. People do
> not stop thinking about multicast just because the SSM work is done.
> 
> 	For all such reasons: Isn't it  appropriate to reserve a
> subrange of the 232/24- address space for SSM?
> 
> 	Regards,
> 	Heinrich Hummel
> 	Siemens AG