[Standards-JIG] the word "multicast" is not appropriate

Carlo v. Loesch CvL at mail.symlynX.com
Thu Jun 1 08:56:31 UTC 2006


Dave Cridland typeth:
| I think it's the best term available, actually. In IP, multicast 
| works by creating a magical address which from that point on is 
| basically handled as a mixture of fanout and broadcast. "Multicast" 
| in XMPP is pretty much the same thing.

The existence of an address doesn't make it a multicast, it's only
a handle. IP has handles for multicast trees, but also for broadcasts
and local hosts. psyc too has handles for multicast trees, like
psyc://psyced.org/@cnet is the handle for the tree that is spawned
off psyced.org for all the recipients of cnet news in the psyc network,
using a tree where a server in australia doesn't have to fetch the news
from germany if there's already a server in australia who has it.
and then there is IRC which does multicasting over its network tree,
so it doesn't have a handle for the tree, as there is only one.

Now we want to introduce handles in XMPP which point to those context lists
of multiple recipients, that's a start for a better distribution, and
it is heading the right way, but as long as there is no tree or mesh
and every server in australia needs to get it from germany, you just can't
call it multicast. You are trying to use a chic term in an inflationary way.

Instead it is a very good idea to create new terms when you want to make
sure people will not presume something to be something else. "broad unicast"
and "smart unicast" make it clear, that you have to read some specs to be
fully aware what they do, and this is a good thing to do, because what those
terms imply has nothing to do with the real meanings of broadcast and multicast.

So anyone coming from any other background will not put wrong expectations
into those words.




More information about the Standards mailing list