All things related to multicast

[Bill Owens <>]

At 8:27 -0700 8/7/02, Kevin C. Almeroth wrote:
> On Wed, 7 Aug 2002, Marshall Eubanks wrote:
>  >>It seems clear to me that SSM will have SAP/SDR like guides, and 
> I wonder that
> > > these have not been created yet.
>
> Yikes!  Why would you need SAP/SDR-like guides?  What function does
> the WWW not serve?

I think the very small selection of multicast A/V content currently 
available makes SDR and similar single-point program guides useful, 
but that in the future they will fade. Today, there's often very 
little interesting content, and just to test multicast it is handy to 
have a single place to go see what people are doing. Of course, the 
majority of announced sources are typically dead, which doesn't help 
when testing!

I do notice, however, that most of the time I find out about 
interesting content from an email announcement or a website, rather 
than SDR. If multicast is ultimately successful in the commodity 
Internet, we can expect many, many sources active at any moment, and 
an SDR with ten thousand entries would be useless (cool, but useless 
;)

As to the issue of ASM versus SSM, I haven't followed the arguments 
each way over the past few years so I can't really comment 
authoritatively. It does seem to me that there are relatively few 
applications where SSM can't be used, but some do exist. Alan's is 
perhaps an example; a large set of data sources that also wish to 
receive each others' data, along with multiple central servers that 
want to listen to all of the data sources. The current system uses 
application-layer multicast with each source opening TCP connections 
to some subset of the central servers, and each of the servers acting 
as reflectors to each other and the other connected sources. This 
allows some filtering and caching, which is handy, but also causes 
bandwidth and server overload. Already there's been a need to impose 
a hierarchy with regional servers interposed between the data sources 
and the central machines.

Of course, multicast for this particular app will not be practical 
for a long time anyway, since Alan is very likely the only person on 
the planet who has a data source that is also on a native multicast 
network - but we can still dream ;)

Bill.