Squeak and Multicast?

Ned Konz ned at bike-nomad.com
Tue Jul 23 17:01:28 UTC 2002


On Tuesday 23 July 2002 09:36 am, Kevin Fisher wrote:
> Thanks Ned, I'll look into this.
>
> Basically I'm just playing around, but the idea of using multicast
> for audio/video broadcasting is intreguing.  It's the way it Should
> Be (TM) done, although unicast seems to rule the day right now.

One of the advantages of Spread is that you can specify various 
different service classes w/r/t ordering and delivery guarantees.

For instance, for typical audio or video streaming, you'd use the FIFO 
service type, which guarantees that all messages will be received by 
group members, and that the messages will be ordered (among messages 
from that sender).

I envision a system where clients can "tap into" an ongoing stream; 
there are issues with re-synchronizing compression, etc., but I've 
noticed that mp3 and mpeg video usually seems to re-synch fairly soon 
(from a human perspective).

Think of this:

program A receives a request from program B (via a separate request 
group subscription) for a video stream. It then begins sending this 
stream via FIFO messages to another group (either a pre-defined one, 
or a specific one requested).

Now, another program C comes along, and queries the current status 
("available channels") by sending a different request for video 
stream status to another group. It gets a response from all currently 
sending video sources, telling it what groups are currently being 
sent video.

It sees that stream coming from A, subscribes to the group, and 
immediately starts receiving messages. These messages are chunks of 
video; perhaps they could be marked every so often as a 
re-synchronizable point in the stream.

Note that A does not have to know anything about any of its 
recipients; they can come and go without affecting A (though it can 
get membership messages so that it can, if it wishes, stop sending 
when there are no more listeners).

-- 
Ned Konz
http://bike-nomad.com
GPG key ID: BEEA7EFE




More information about the Squeak-dev mailing list