Re: Auto-tunnel Rant

[almeroth@cs.ucsb.edu (Kevin C. Almeroth)]

Well, I'll add my voice to the rest of the noise.  As I see it,
auto-tunneling is fine.  Why?

1.  Let's not get in the business of protecting people from
    themselves (in reference to Dino's comment about large
    fanout is bad...  it is, sure, but who cares?)

2.  Let people run whatever they want in their own cloud.
    To force people to only do network-layer multicast is
    wrong because:

       (a) You should be able to do whatever you want (See 1.)

       (b) We are in a transition period and not every 
           single device supports multicast.  Until every
           single device can handle multicast we need things
           like tunneling (any maybe auto-).  Why?  See 3.

3.  An infrastructure that has devices that don't support
    multicast and so no way for the eyeballs on the other
    end to see the content is ultimately defeating for
    multicast.  Yes we can be multicast advocates, but saying
    that someone can't play in our sandbox because they 
    haven't bought the latest equipment from Vendor X hurts
    us more than it hurts the rest of the world.

In the end:  who cares if there is tunneling?  or even
auto-tunneling?  Who cares if people expend effort deploying
non-network-layer multicast?  If they weren't smart enough in
the first place they probably wouldn't have been smart enough
to deploy multicast anyway.

NOW...  the real question is should the IETF be standardizing
a way to do this?  My intuition...  NO.  Let some clever vendor
sell a piece of hardware or software that makes it mostly
seemless.  Done.  It creates pain that is solved by...  
ta dah:  network-layer multicast.

-Kevin